S12E4: The New Developer Experience with Factory Co-Founders, Matan Grinberg & Eno Reyes

And I'm just ready for everyone to be like, wait, this is like the most exciting time to be a developer.

Yeah, we're probably at the lowest number of developers there will ever be for the rest of human history.

Welcome to The Room Podcast. I'm Claudia Laurie, the founder.

And I'm Madison McElwain, the funder. We are on a mission to open the door to the room where it happens, sharing the stories of some of the world's most iconic builders.

Authentic storytelling is core to our community. It all started as two best friends who met in the heart of San Francisco as product managers in 2018. When I left Uber to start my company, Madison was our first yes. It was these conversations that birthed The Room Podcast.

And nearly 5 years later, our guests have built businesses with a combined enterprise value of over $100 billion. You'll recognize conversations with your favorite technology founders from Perplexity to Vercel to Zillow, to Cloudflare, all across our 110-episode catalog. The best way for you to get to know our community, both online and offline, is by subscribing to the podcast and our newsletter, which you can find at theroompodcast.com. Okay, let's open the door. Before we dive into today's story, we have a few short messages from our sponsors. Perkins Coie supports the most innovative entrepreneurs and investors in fast-moving and high-growth sectors. Addressing their myriad of legal needs. But the firm doesn't just provide end-to-end legal and business counseling to its startup clients. It also facilitates introductions to key advisors and sources of capital. Perkins Coie's interactive website, Startup Percolator, offers access to programs, resources, and rich dynamic content designed to assist entrepreneurs on their startup journey. To learn more, go to startuppercolator.com. .com and perkinscoie.com.

This podcast is brought to you by Mercury, the fintech trusted by over 200,000 businesses, including The Room Podcast, for banking, credit cards, and financial workflows. The truth is banking can do more. Mercury brings everything you do with money into one product that feels extraordinary to use. Visit mercury.com to apply in 10 minutes for banking that makes more possible for your business. Welcome back to The Room podcast. Today we're thrilled to take you inside the room with co-founders Matan Grinberg and Eno Reis of Factory. Just a year after ChatGPT burst onto the scene, Factory set out to become the command center for software development, unifying all engineering context from GitHub to Notion and empowering product and engineering teams to collaborate with autonomous AI systems. With backgrounds in physics and computer science, as well as experience at companies like Microsoft and Hugging Face, Matan and Iino bring a fresh perspective on how AI is reshaping the role of the developer. Their journey has already attracted $20 million in funding from Sequoia, Lux Capital, and Mantis Ventures, positioning them at the forefront of the AI-native movement. In this conversation, we dive into their early influences, the aha moment that sparked Factory's creation, and the challenges of convincing dev teams to let AI execute code. We also explore how AI is shifting go-to-market strategies, transforming collaboration in both startups and large-scale organizations. Innovations and the fast evolving landscape of LLMs and foundational models. Key topics in this episode include the future of AI-driven development, fundraising in the age of GenAI, and how Factory aims to stand out in an increasingly crowded market. Both Madison and I were so impressed by how these young founders are thinking about the space, and I know you will be too. Let's open the door.

Well, first of all, Matan, I know, thank you so much for joining us today in the room.

Thank you for having us. Yeah, thanks.

Excited to be here.

We're super stoked to have you in person in the physical room. So let's get started with you guys. Where did you both grow up and how has that shaped your view of the world? Matan, I'm going to you first.

Cool. Yeah, so I grew up in Palo Alto, California. Well, originally moved around a lot in the South Bay, but I choose to define Palo Alto as where I grew up. Sounds good. And hated startups.

Oh.

Wanted nothing to do with them just because every— in Palo Alto, every single good idea is, oh, you should make a startup for that. I, you know, I think really didn't like that. So I was like, I'm gonna do the exact opposite thing, whatever's the furthest from startups. So decided to get into theoretical physics and in particular string theory. And so set off for basically the next 10 years to focus on that, which is what brought me to Princeton, which is where I first, you know, got connected to Eno. And after that, did a master's at Cambridge in the UK, came to Berkeley to do a PhD. At Berkeley, I kind of got there at the same time that Eno ended up coming to San Francisco. So I'll let him give a little background before I explain how we then joined forces.

I want to jump the gun on the meet-cue for you guys. So let's keep going with Eno. Like, where did you grow up and how has that shaped who you are today?

Yeah, absolutely. So I grew up kind of on the opposite side of the US in Atlanta. So Southeast. I— my parents both met at Savannah College of Art and Design. They were like graphic design and illustration. So I grew up in a very like art-forward household, but also very technology-forward. So like Photoshop and all these kind of creative tools were like easily accessed for me. And so a lot of my early exposure to technology was like creative technology. And so I think that growing up, to me, it always felt like technology was this tool for like self-expression, less so on the like software, like B2B SaaS kind of side. That happened later, I guess. But, It definitely felt growing up like the thing that was always encouraged to me was think about like the future, think about what technology can unlock for creative expression. And then went to Princeton, did CS and like cognitive science. So a lot of the focus there was on neuro and kind of trying to approximate how brains might think but with software. And that I think really put the itch in me to like work in AI.

I've been saying that the internet feels fun again for the last month because there just is this zeitgeist moment of creativity that's happening. And a lot of what you're building with Factory is empowering that within the developer workflow. So we're going to get there. But before I jump the gun, you already mentioned you did not see yourself being a founder, but, you know, did you always think you'd become a founder?

I think that the idea of starting something and building it has definitely been on my mind. In college, I had a a creative technology agency. So it was like 10, 15 people, mostly contractors. And then what my roommate at the time— so it was like dorm room startup. But this was not at all like a venture-backed endeavor. It was more like an LLC that wrapped around consulting, basically. And so I think that that definitely put the itch in me.

Cool. Well, you mentioned Princeton. It's where you guys met. Was it inevitable that you guys would start something together?

So it's funny you mention that because— so we both went to Princeton. We were a year apart. And we had what might be like the maximum overlap of mutual friends without ever having spoken to each other.

Perfect.

Like, I think actually like 150. We, we've talked about this so many times where like, yeah, his best friend was like really close with my best friend, yet somehow we never had like a one-on-one conversation. We probably went to like so many of the same events, like we recognized each other, but just never a one-on-one conversation.

Yeah, we were definitely like in a, we were at least one class that we know of where we were like probably sitting right next to each other.

Yeah.

So how come we're here today in this room? When did that happen?

Yeah.

So, so So, you know, Eno found his way to San Francisco by way of Seattle.

Love it.

I found my way through. Exactly. I found my way through Berkeley. And just maybe a couple months after ChatGPT came out, there was like a frenzy of hackathons in San Francisco. Every other day there was like another hackathon. You know, it was a big thing if you got invited to this one or that one or whatever it was. It kind of died down, you know, shortly thereafter. But at one of these, you know, we were both there. And saw each other from across the room, and it was like, wait a minute, like, I remember you. We both like kind of looked at each other and recognized each other. And, you know, we ended up going and saying hi and kind of reconnecting. And that ended up being 8 days before we started Factory.

Nice.

Wow.

Which was a pretty— it was, you know, one of those things where I was like, we like to joke that it was intellectual love at first sight. You know, when you know, you know. And yeah, so that was, that was how we reconnected.

Yeah, I, I was like in a moment where at Hugging Face we were just talking to a bunch of companies, like very like large enterprises that all had no idea what to do. They were like being told AI is here and they were coming to us saying like, hey, please help us. And it was very obvious. So they were sending me to all these hackathons to basically talk to people. And, and so seeing him and being like, oh, that's somebody I actually know at this wave of just random people. Was so reassuring. And technically at the time, I had actually started talking to this other guy about starting something maybe. Instantly, once I met Matan, I was like, no, no, no, Matan is the guy for sure.

Love that. Love that conviction in each other. And you're living in the world of theory, as you mentioned, doing the research on the cutting edge. And you're working every day in the developer toolkit and workflow. What were some inefficiencies either both of you were seeing or you particularly as a developer were seeing kind of in your day-to-day jobs?

Yeah, I think that the thing that was definitely most evident was at the time, the like strongest conviction tool for developers to use AI was autocomplete. And it just really felt like what we were seeing was an evolving paradigm of, well, what if we put these LLMs in a loop? And what if we let the LLMs like you know, call functions. The word agent really hadn't, at least in AI circles, had the prevalence that it now has. And I think it was you that when we were talking, Matan was like, have you ever heard of this baby AGI thing? And for me, that was like a repo with like 50 stars that a coworker had mentioned one-off. And I was like, well, how do you know about that? And I don't know.

Yeah. And so, and one thing I remember, so this was also, this was way before people were going crazy about agents. This was also, you know, like I said, a couple months after ChatGPT came out. And something I remember being particularly struck by was like within 2 months, Eno and the way that he developed, like solved, the way that he wrote code was completely different. And this is something that to this day I still am in awe about is like the second there is a more optimal way to do something, Eno will be the first person to just like fully adopt exactly how to do it. And one thing that like came to mind was the way he was doing it in the, this was in like I guess April of 2023, maybe March, was, you know, you have your IDE and then you have ChatGPT and you're like swiping between those. And clearly that is not like the optimal interaction pattern for, you know, building software with AI. And so this was something that was like, these tools are only gonna get more powerful. Clearly the optimal interaction pattern is not there at all. And like no one is actually thinking about how do we, like in production settings, change the way that people develop. And so that was what, you know, we got, got carried away with that. Yeah.

So this is an incredible backdrop for us to discuss Factree in more details, but I'm curious, there was basically an 8-day window between intellectual love at first sight and Factree being co-founded by you two.

Yes.

What happened during those 8 days?

Ooh.

Yeah.

Yeah. That is a, that is a great question. So, uh, similar to how Ina was talking, uh, to, to another person, there was a friend of mine who I was talking about. So I was at Berkeley doing string theory, like nothing to do with AI or software engineering. The only thing it has to do with AI is that it's like involves math and it's easy to then apply it.

Yeah.

Yeah. And a piece of insight that this friend gave me was, hey, you know, while you're a student, you should email more people because people are much more likely and inclined to help you if you're a student because it's like you're not looking for it. It's just advice. And so I'm looking and I'm thinking like, okay, you know, who are the people that I could reach out to that I would be even curious to like get their advice about? And there was this randomly in my YouTube recommendations, there was a podcast recorded during COVID So it was like an online, not, not, not face-to-face. There's a podcast with this VC who I recognized his name because he also used to be a string theorist. And in fact, one of the papers that I had published at Princeton cited one of his works, which there's not really a string theory to VC pipeline. So I was, I was very intrigued by that. Ended up sending him an email. It's Sean McGuire from Sequoia.

Nice.

Sent him an email, was just like, hey, I used to do string theory like you. I published a paper with this guy who's very famous in string theory. His name is Juan Maldacena. If you're not in string theory, you probably don't know.

I'm sorry, I don't know that influencer.

Yeah, yeah.

He is arguably the most, like, prominent string theorist. So it's pretty crazy for an undergrad. I don't— he basically does not work with undergrads. So the fact that Matan did definitely caught Sean's eye.

Yeah. So, and I called him Juan too, as like an if you know, you know, like you don't say the last name. And Sean replied. We ended up meeting down in Menlo Park. It was, it was supposed to be like a 30-minute meeting. Sean loves walking. We ended up walking for 3 hours. I was in— he was in running shoes. I was in like random sneakers. I had so many blisters by the end of that. But at the end of that meeting, he was basically like, you should either join Twitter right now because Elon just took over. And if you choose to join, that's badass. Or 2, you should join one of my portfolio companies, or 3, you should start a company. And so at the time I was thinking about like Code Generation was really excited by it. And I think, you know, and I had that love at first sight the day between emailing Sean and when we actually met in person. And so like the coincidences there were ridiculous. I remember the next day, you know, and I got coffee at noon, like in the middle of the week. And I was like, hey, like, you know, what do you know about code generation? He was basically like, a lot. And I was like, okay, great, let's do it. And, you know, for the next 72 hours, we ended up working on, retrospectively, a really shitty demo. But, you know, it was, at the time, I think it was really, really exciting. Sean, I emailed him and I was like, hey, I think there's something cool I want to show you. He's like, okay, like, whatever, let's hop on a Zoom and see. Showed it to him and he was like, okay, like, this is all right. And I was like, oh, like, just all right? Like, I think it's pretty cool. He's like, okay, would you work on it full time? And I was like, yes. And he was like, okay, drop out of your PhD and send me a screenshot.

Oh my gosh.

Wow. And so didn't tell my parents, dropped out of the PhD, sent him the screenshot. He's like, all right, you guys have a meeting with Sequoia tomorrow morning. Like, be ready to present a demo and a deck.

Shut up.

And, and, and I'll tell you, putting together a demo of an AI system building code, like an application from scratch, 2 years ago, like basically a couple of months after ChatGPT came out and then being told like, well, I guess we kind of knew. So we knew what we were building for, but like knowing that we would then have to present to like the longest table of VCs that you've ever seen in your life was, was quite a motivator.

So you guys show up in pajamas?

No, not quite. It was definitely a polished, polished presentation. But the I still remember during that demo. So Matan is kind of like, I'm on my laptop, I have like logs and error things. On his laptop, you just see the demo. And there was, we knew, LLMs are non-deterministic. We knew there was like a 1 in 5 chance that it breaks on this one step. And so it was effectively this dice roll as it crosses this threshold of like success or not. That was basically like 80%, which we felt was pretty good. Yeah. But 20% chance that it just basically, you know, popped off and didn't work.

Don't leave us hanging. Did it work?

It worked. It worked.

Yeah.

And you're here today.

And we're here today. It was also one of those cases where I think it was helpful to not know anything about like VC or that kind of world because like retrospectively now it was a table of people that I probably should have been intimidated by. But at the time it was like, I don't know anything about VC. Like, whatever, let's just have fun.

Yeah.

And yeah, got a call from them the next day and Factory was born.

We'll dive a little bit more into your funding journey in a moment, but I'd love to talk about that demo. What was in the deck? Factory is the command center for software development. For folks that might be less technical, tell us, what does that tactically mean?

Yeah, so I guess at the time, the key focus for us was like use case specific agents for software development. We called them droids. In fact, Fun fact, we were initially incorporated as the San Francisco Droid Company.

Sick.

Upon getting some advice from our legal team, the fact that Lucasfilm is apparently quite litigious, we decided to change the name.

Shoot.

Yeah, to the San Francisco AI Factory. So, but yeah, so at the time it was, you know, agents focused on particular steps of the software development lifecycle. So like understanding your code base, you know, planning your project, actually doing the code, writing tests, reviewing, documenting. And how that's shifted over time is that we've realized that the alpha isn't necessarily in the domain-specific agent itself, but rather providing agentic AI workflows to developers in the enterprise in these complex settings. So being able to unify different sources of context that an engineer might need. So like, their code, but also their Google Docs or, you know, their Jira tickets, things like that. And bringing those together with agentic workflows so that they can collaborate with this agentic AI when they'd like to, but then also delegate tasks that they don't even want to work on at all.

As a startup founder with a very small team of engineers, I resonate with how efficient that would be as an improvement to a developer's workflow. But a lot of times startup engineers, even big company engineers are not that comfortable with adopting a new tool in sort of their development process. I've certainly sent, you know, our engineers tools and they're like, oh, this looks a little sketch, like whatever. What was the biggest challenge around getting developers to use Factory? And how has that evolved over the past few months?

Yeah, no, I think that this is, and this has definitely evolved a ton. There's a clear need for trust with AI tools. And I think the biggest thing is giving developers the ability to, as Matan mentioned, both collaborate and delegate with these systems. Because if you just say to something, I want you to do this task, and then it runs off and comes back to you, that was actually basically the original version of the droids. It's like you said, I want documentation. It worked and it came back to you. And I think that there's this kind of bar that gets set where you really need to be almost 100% successful or else you lose trust. And so then you start thinking, well, what if we showed them the steps that it took to get there? What if we showed them what are the options that they have from the beginning? What if we showed them that they can interrupt and interact with these systems while they work? And you start to realize that there's kind of a, a set of different developers, some that are very willing to drive that whole process themselves. And those are the ones who are a little less trusting of AI. There's a whole other set that is, you know, like, let me, let's YOLO it, let's send the agent off, let it come back. And I want to see those results. And so building out the platform to respect those, like, transparency and decision-making processes was probably the biggest step change that has led to the adoption it now has.

Yeah, that makes a lot of sense. Transparency, trust, really key sort of design principles here. I was recently talking to one of our engineers around just sort of like the hiring market right now. New grad engineers are finding it harder to find jobs because AI has really accelerated, you know, more senior engineers' workflows. In your opinion, like, how is AI shaping the role of developers, engineers, you know, in the next 5 years?

Yep.

No, go ahead.

So I think one thing about this that Eno said in the office the other day that I loved is that the center of gravity of software development is changing where it used to be. And obviously, you know, physics background.

Yeah, yeah, go on.

Yeah.

He teaches a lot of those. Yeah, exactly. But the idea that, you know, originally the focus was on coding, but as these systems get better and better at coding, if you give them clear and concise instructions, that means that the emphasis is now shifting towards understanding and planning. So what that means is, you know, as a developer, you are spending more of your time making sure you're very clear about the certain constraints that you want to satisfy with whatever it is that you're building. So that's kind of on the like individual developer level. What that means in terms of like, you know, organizations as a whole. And I think a lot of people talk about, you know, developers being replaced, which we are very adamantly— I don't even think that like, it's not even that it's like bad to say these things. I also think it's like categorically false. It just means that the table stakes for what an organization with 10 developers should be able to ship now grows. Yeah. It would be true if like one company had access to AI and none others did. But the fact that everyone has access, it just means like now having like shipping 5 different things in one month is like expected of everyone, right? Yeah.

So yeah, that's kind of the, the And now with AI just being so table stakes to every, you know, software business, um, you know, you have to go after these customers and there's often kind of a fight to win some of these large deals. Tell us about your go-to-market strategy and tell us about how you won your first, like, hallmark customer.

Yeah. So I think something is talking about the go-to-market. So we're generally, we're very focused on the enterprise in particular. That's probably the least served right now because it's really fun to, you know, show demo videos of like, look, like my 7-year-old just, you know, created this app from scratch and like, that's really fun. It's really exciting. It's also, I think, just great for, you know, less experienced developers to now be able to do these things.

Totally.

But in the actual enterprise, that's a very, very different story, right? Like you're not building apps as zero to one. Um, you have 30-year-old codebases. Half of the code might be dead. You have out-of-date docs. You have parts of the codebase that, you know, the last person who touched it is like retired. And, you know, the considerations for you are not I want to build this thing 0 to 1. It's like I want to make sure that I can change this thing and I don't break everything else that relies on it. That's a lot less sexy, a lot less conducive for demos. But the value that you can deliver if you do that is way, way higher. And so, That's what we're focused on. Because of that, a lot of the companies that we go to, they actually don't have, like, even though it seems like there are a million code generation tools, a lot of the enterprises actually don't have people going to them because it's just not as fun and viral of a use case.

And it's a lot harder. It's just there's a barrier of security, of thoughtfulness, and where's the data located? Are you training on these? What is the minimum amount of spend of tokens that you can use. I think a lot of tools just aren't built for that conversation.

Yeah, absolutely. And so in terms of the first enterprise, so actually the kind of first from originally from the Droid paradigm of things when we were focused on these particular use cases, it was actually the code review Droid that people really, really liked. The, and it's still to this day, it's something that, you know, all of our customers use, which is as you have more code generated by AI and you give each developer the ability to produce more and more code every single day, you also need to increase your standards or your measures to verify that the code that you're shipping to production is actually good code. And our review product helps speed up that review process. And so that was like the initial wedge with our first enterprise customer. They're still using it to this day. We actually just expanded within their org to a lot more use cases. Yeah. So that was a lot smaller. That's a, that was a 50-engineer organization. We're now working with orgs that are, you know, orders of magnitude larger. But I still remember the feeling of like when they actually converted. That was so indescribable. Like when you spend so much time and the team is like heads down grinding for so long and then you get that, you know, I remember on that call, with, you know, our point of contact, who was basically like, yes, by the way, I'm going to make the recommendation to my boss that like we procure. Yeah, I remember being on the verge of tears, which was— I'm not typically, you know, particularly emotional. And that was, that was— yeah, it was, it was pretty crazy.

Yeah, I think that the, the, the strategy that we had early on was effectively just go small and then go bigger and bigger. Which I would recommend. I think that actually does work pretty well if you're like a first-time founder. You want to be in the enterprise market, but you don't have any of the things necessary to do that. It's much better to like get those shots off with low-stakes customers. I think that now that once you get that first, a lot of the motion just becomes learning how varied enterprises are. There really is no one-size-fits-all. There are entire countries' worth of companies that have completely different requirements, different compliance, different preferences. And so I'd say that a lot of the time that followed was just iterating with these companies, trying to figure out what do people think about CodeGen? There are IDEs, there's agents, there's things that are kind of like somewhere in between, and talking to them and saying, what is your AI strategy? What are you looking for? How do you think about this? And then adapting.

No, even just this conversation is helpful. Claudia and I were prepping for this, of course, reading all that we could about Factory, and we were kind of like, yeah, how do we square this, right? Like, where does it fit? Because you're integrating so many important parts of the existing developer stack. I mean, we talked about Jira briefly, and then you have Linear now, and then you have your existing IDEs, etc. So just like understanding where you even fit in the market, it sounds like you guys are creating a category.

That's right.

I think so, yeah.

Absolutely.

Am I next on questions?

Yeah, you're next.

Okay, okay, perfect. 'Cause I was gonna talk a little bit about the fundraising conversation. You know, we already answered the question, who was the first person to say yes to you? Sounds like it was Sean at Sequoia.

That's right.

And as you guys went that, first of all, step back, congratulations. Huge congratulations. Yeah, amazing $20 million round, $125 million valuation. You know, you have amazing partners. You have Sequoia, you have Lux Capital, Mantis. Have you guys met the Chainsmokers?

Yeah.

Yeah.

Vibes?

What do we think?

They're great. They're great.

Fantastic. They came by the office to talk to a bunch of the team. One of our, like, our head of biz ops, Fran, that you met, she's awesome and she, you know, knows them very well.

Of course. Of course. Absolutely. Well, that's really fun. I love the Chainsmokers. I'm not going to lie. I guess just for the AI ecosystem, right? We have a lot of founders listening who are hoping to be in your shoes. What advice would you have for them on going to market with a Series A fundraise?

Yeah, that's a, that's a good question. I mean, I think, uh, importantly, I think what people can really tell immediately is if a team views fundraising as a goal or as like a tool to get to goals. And I think that's something that we've been extremely disciplined about. And I think people tend to appreciate that because that kind of shows longer-term thinking for the business, especially in, in our category. I think cogeneration is one that definitely has a lot of money being put into it, which I think is a great thing. This is a very important space. But understanding what the utility you actually get from doing a fundraise, knowing that you're giving away a big piece of your company to do so, and making sure, you know, you do these things when you have optimal leverage, when you actually need it as a, you know, as a business, I think that's really important. And yeah, I think doing as much as you can to remember that like the zeros do not matter. Like it's all the, like those headlines are fun. They don't matter. Fundraising announcements do not matter except to whatever, help hiring or help, uh, you know, top of funnel. It does not matter. Like the thing that actually matters is the product, the go-to-market. That is the thing that will actually build a long-lasting business.

Yeah. I think that the, uh, pretty aligned with what you're saying. If you set out your goals for your company and you, like, it feels like if you're building SaaS that actually does scale with the, just the actual software itself rather than with like some hard cost, obviously totally different if you're building robots or some physical world thing, you will kind of naturally find those points where you probably have to fundraise to move on. And there's actually not many of them. It's mainly, you want to grow very quickly, you've figured out a sales motion that actually works, there is some major research that you probably need a lot more people than you currently have. I mean, other than that, I would really recommend looking at the space, at all of the different players that are raising, that are involved, and trying to figure out like which of those paths seems most like the one that I'd like to follow.

And to add on, I think also making sure that it plays to your strengths, right? Like if there's a category where to even enter you need to raise $1 billion and you don't have a strong network, then like, you know, probably don't spend a lot of time doing that or focus on building that network first if that's the thing that you're, you know, set on doing.

So I'd love to hit on strengths related to, you know, your background with Microsoft and Hugging Face. You know, you started your career building within this ecosystem. And so for you all, I'm curious just with that POV as a developer, like What does it mean to be AI native?

Yeah, yeah, that's a great question. I think that the— this really starts with how you think about running a team that builds software. So like we build software for teams that build software, but we're also a team that builds software. And so like for us, we use, you know, our own tool all the time. And in fact, I think something— I don't remember the most recent stat, but it's somewhere between like 85 to 90% of our codebase is generated with older versions and our current version of our tool of the Factory platform. And so you can kind of sense this if you were to just shadow anyone from our team. All of our notes are transcribed automatically. I do one-on-ones transcribing. I bring them into our platform and I have droids help me organize tasks, next steps. A lot of our pod organization of the different engineering teams is basically built around the minimum number of people who can collaborate in our tool together and solve problems. And we also have a much more horizontal kind of structure where lots of developers have the freedom. And I think this kind of gets to a bit of that earlier question of how is like the role of a developer changing? We really see there being one type of persona, which is a, like, somewhere between product engineering and eng technical management, where with tools like our platform and many others, you can manage things like tickets automatically. You can organize work based on high-level company principles. And so as long as you have as a company, like, really tight definitions of what it means to build product, what it means to do engineering at factory, AI tools and humans who are competent enough can actually be the product manager, the engineer, the implementer, and the QA. And so a lot of our team is, you know, members of technical staff because they do everything, which is really exciting.

Yeah, I love that vision for it. And I think it's helpful for our listener base because everyone's being pushed to be AI native. But the reality is most companies are barely digitally native. And so as you all are thinking about incumbents in the space, and I'm thinking more like, yeah, kind of like the Jiras of the world that have just been sitting in this tech stack for a very long time where, or frankly, what does Excel that's helping people with some of these tools and managements. How do you think about competing when no doubt the incumbents are trying to make themselves as AI native as possible?

Yeah, yeah, that's a great question. I mean, I think a big thing here is understanding the end user and understanding what their existing workflow looks like. And, you know, balancing on one hand, we are very intentionally not iterative in how we think development should change, right? Like there are, there are tools that are adding AI onto existing workflows. So that's like AI IDEs. Developers are already in the IDE. Let's add AI on top of it. Boom. Like now they're using it. Um, something we talk about internally a lot is the, the Henry Ford quote, which is, if I ask people what they wanted, they would say faster horses. Um, and we are very much, you know, trying to build the automobile, not faster horses, which is harder on one hand because developers can be particularly stubborn. And the IDE paradigm has existed for like 20 years. Some people still, you know, use Vim or Emacs. And, you know, you're going to potentially one of the most, you know, stereotypically stubborn groups of people and saying, we are going to, you know, provide this tool in a net new workflow. So that kind of increases the barrier for entry. But then at the same time, you kind of get to start with this blank slate and this future of software development. And software development is unequivocally changing. Dramatically over the next 5 years. We get to show them what this new way to build software will be without kind of the baggage of the IDE and the preconceived notions that they might have there. Now, given that, we also do need to meet the developers where they are and understand that it's a little unrealistic to, let's say, go to a 1,000-engineer organization and say, all right, overnight you guys are now going to do everything this way. So having what we call inlets into Factory is really important. So if they have, let's say, Jira tickets, you know, giving them a link on every engineering Jira ticket that they have that brings them into Factory with all of the relevant context. That's something that we've seen as a huge, huge way to drive usage and adoption within the orgs that we work with. Or similarly, during the PR review process, you know, if there's an issue that pops up in a review, they might want to go to Factory, but if they have to go and like take the link and bring it in, it's kind of a pain. So we have linking that allows them with one click to go in and now be in a session in Factory that has all of the relevant context. And that's another way to kind of like win over a new user, make them realize, oh wait, you know, there's a lot more that I can do here.

Yeah, I think that the— and this is such a great point of like, what is the balance between when you're defining a new category, meeting people where they are and saying there might be just a better way to do this? I think it really required us to— and I very much dislike saying the word first principles. It feels way too like startup-y.

You're forgiven.

But, but it kind of does require you to rethink from first principles what actually matters when you have an agent that can write code, test that code. It knows where to find the information that, that it needs to, to solve the problem. Most of it, the challenge is directional. It's like, I want you to set these constraints. I want you to go in that direction. And so there are certain aspects like IDE is kind of a catch-all term. It's just integrated development environment, right? That could be anything. It's some piece of software that runs on your computer and maybe it has a terminal so you can run commands. Maybe it has a file viewer. If you look at how different tools are evolving, the interfaces are starting to be a lot less the actual old IDE stuff and a lot more the agent and the, the tools that you can view it. I think what a lot of the players in the space are gonna learn are some of the lessons that we've been kind of working on already, which is What matters is showing what the agent is actually doing, showing the information that the agent has in its head, showing the steps it needs to take next and what it's going to do next, and letting the user have control to a very fine-grained degree of every little one of those things. And when you do that, you build something that doesn't really look like an IDE. But as the agents start to get better, it starts to become probably the new form of how we'll build software.

And to that end, I think something that's, you know, sticking with our horse-car analogy, it's like, in theory, can you iterate your way from a horse to a car? It would look pretty messy. I don't know, like Ship of Theseus, like, you know, remove like a leg, put a wheel. But yeah, I think it's— I think it might be better to just, you know, start over, start from the engine.

Yeah.

Yeah.

So, well, you guys are building in a rapidly changing environment and you've carved out this incredible category and you're working towards redefining the developer workflow. But of course, you're being built on top of LLMs and It's been a really fun 2 years watching the larger LLM providers come to market. Unfortunately, they've named one after you with Claude. I don't know what to do with that. It's like the new Siri. It's just getting more and more confusing. But, you know, all jokes aside, AWS is making some really big bets on kind of this competition model. They're launching Nova, their own model, but they also invested in Anthropic. So it's shifting under your feet as building on top of these LLMs. And so we're curious how you're thinking about the model landscape.

Yeah, I mean, and this is, this is— coopetition is a hilarious word, by the way, that I think that the— because we work with AWS, we work with Azure, we work with Anthropic, OpenAI, the XAI team, like we work with a bunch of different model providers, cloud providers. And at the end of the day, I think that there is this kind of like loose sense that everyone is genuinely interested in progressing this whole field. So our partners work really nicely with us. They give us early, early access to models. We— our research team gives them evals and feedback on things. It's very close, tight cooperation. Of course, everyone involved in a multi-trillion-dollar space like software development wants to build AI software development tools. I do think that there is, you know, a responsible amount of, you know, we're doing this and we're probably not going to tell you about that. That happens from everyone. but on a daily basis it feels pretty good vibes.

Yeah. And I think another thing that's important is, so one, from our perspective, the thing that matters most is just giving our customers whatever model is the most cutting edge of the day, which honestly changes on a very frequent basis.

Yes.

Which is why it's so important to have these close partnerships so we can get early access and make sure we modify things such that, you know, when they do want to use, you know, let's say Grok 3 or Sonnet 3.7, that the prompts are adjusted accordingly so that from their perspective, they just keep, you know, going about what it is that they're doing. They just change the model and, you know, things start working better.

Yeah.

So that's one thing that's important. Another is this is why we're really focused on the enterprise, because for use cases of like 0 to 1 app building, whatever model is best is going to be the best at that because a core competency for all of these models is code generation. Now it's different when you're actually deployed in a very large enterprise where Making the model better at code doesn't mean you can suddenly just like drop it into a 30-year-old codebase that like 50,000 engineers have been working on. That's where the differentiated value that we provide gets the largest. Whereas, you know, the zero-to-one, like help my niece build a fun app, that's, you know, not really, not really where we're trying to go.

So yeah, I know that's helpful to frame it up. And that just makes me want to ask this question almost a layer deeper in the stack, pun intended. Like, do you guys care about chips at all? Are you thinking about NVIDIA versus Trainium with AWS? Like, is that coming into your thought process today?

I would say that only in the sense that like we do work a little bit more closely on like some like reinforcement fine-tuning for very specific tasks. It's definitely not like the biggest thing that we do. And so there are certain decisions that have to sometimes be made. Which provider do we use? What chipsets are available to those providers? I wouldn't say that it's very top of mind, though, because most of the inference, probably like 90 95, 99% of the inference through our platform happens with big models, big LLMs on existing platforms.

Most of the interest there is as either a fan or like, you know, for broader geopolitical interest.

Yeah, well, it's fascinating.

Yeah, exactly.

Yeah, yeah, yeah, yeah. No, I'm kind of obsessed with seeing how it all plays out. So, so we'll keep it, keep it, keep it going. Well, you kind of mentioned some fun jokes around the no-code platforms, the v0s, the Replit, Lovable. I'm probably similar to the nieces building on those platforms, which is a good thing.

I think it's, I think it's a great thing. I like, I think it is good for truly everyone. Yeah.

Yeah. You mentioned something earlier that you feel like the internet is like fun again.

Totally.

I think that there was an era where Flash games were everywhere because it was so easy to make a Flash game, so easy to, to just build something that was creative. And I think that we see that now because of, you know, Lovable, Replit, like for those tools, if you just want to build something, that's awesome. And I, it's like the dream of technologists.

We have new hats dropping soon that say Vibe Coder on them. Nice.

Actually, we're debating vibe coder versus vibe coding. What do you think?

I like the label as opposed to it's like I identify—

As a vibe coder.

Yeah.

I truly do, by the way. I started my career as a product manager and I had to learn what an API was on the job. I mean that seriously. I was a history major, engineering minor, but design thinking, so nothing fancy. I just think it's been so cool to see the evolution of even in my own self, my confidence in some of these platforms. Maybe taking a more macro question for you, Where do you guys see this going in 5 years? Like, what is Factory in 5 years?

I think that's a really good way to kind of lay the question up, because I think something that's really exciting about whether it's the more like vibe coding or enterprise setting is like the broader theme here that's happening is basically lowering the barriers from idea to bringing it to life, which is really like at a high level, it's kind of a more beautiful manifestation of what capitalism should be. I know that sounds funny.

No, no, let's go there.

Right.

Like right now, if you have the best idea, Like it is in America, it is in our interest to bring the best ideas to life. But if you have the best idea and you are not in a privileged background where you have access to hire 30 engineers or you have access to raising VC capital, we have these latent ideas that are good, that are just kind of like sitting there unrealized. And with tools like this, you now can actually start realizing those ideas without having access to some of these things. Yeah. And that's a net good because that just means better ideas will come to life more quickly. And again, that might start in the realm of vibe coding and then work its way up. If it's such a good idea to like an enterprise setting, perhaps it might end up being, you know, an enormous company that starts from, you know, someone who might not know engineering and is just vibe coding. And then it gets a lot of attention and then becomes much larger than that.

Yeah, at risk of sounding like an economist, I think what you're laying out is a really interesting shift in the GDP equation because people are making assumptions that if we increase the amount of engineers we need to produce code, We're going to lose GDP, but we're going to gain productivity. So then what else happens but then bring in innovation?

Yeah, absolutely.

And I'm excited by that framing. And you guys are at the forefront of empowering those teams and also just being creative. It sounds like your whole career started with creativity with your parents.

Yeah, absolutely. And I think the creativity, the exploration is undoubtedly already happening. And I think there's also this other— exact opposite end in scale. There's no way that we can colonize Mars, build humanoid robots, cure diseases without a, like, order of magnitude more code being generated, an order of magnitude more productivity per individual. And so I think that there's this kind of slightly less explored idea that it is unlocking complexity in the problems that we solve that just would not have been accessible before. And that happens, I think, a lot in the enterprise setting where a company that already has the resources, the scale to unlock like capital to inject into a problem can now solve very, very substantial problems, which I think is something that a lot of our partners are in finance, healthcare, you know, large B2B SaaS and their use cases are crazy. It's like we might cure a major disease, we might redefine financial markets in an entirely another part of the planet. And for that to see that being accelerated by our platform has been like unbelievably cool.

So are you hiring?

We are indeed.

Okay.

Indeed.

Perfect.

We will put a link. Yeah, put a link.

Killers only.

Absolutely. Vibe coders, you can think about it.

Yeah.

No, we accept vibe coders.

All of our coding interviews, you're allowed to use any AI coding tools that you'd like to.

AI native.

Yeah, absolutely.

We love that.

Well, we've talked a lot about, you know, the founding journey of Factory and, you know, you guys have accomplished a lot in a very limited amount of time. So I'm sure what Factory looks like at the end of this year is very different from where it is now. You know, what are you most excited about on Factory's roadmap?

What am I most excited about? Probably like O9 from OpenAI or something like that.

No, no.

But in reality, I think the thing that's probably most exciting is we now see people delegating very substantial tasks to our agents and solving them end to end. And so there's kind of this open question of how do you slowly transform a product that is indeed meant for individual developers to allow teams of these systems to work together and bring humans in the loop? And so there's a lot, I think, of like practical insight that you get from seeing these agents just work end to end. People will just tell you, hey, it would actually be great if my colleague could jump in here and do this, or hey, it would be great if these two agents could talk to each other. And so there's just a lot that goes into having very competent systems working independently.

And Matan, what about you?

Yeah, I think something that I'm most excited about is as we do more and more of these deployments to more high-profile organizations like Fortune 500s, the narrative— I cannot wait for the narrative to change. Currently, the narrative is like, you know, AI is going to replace developers.

Yeah.

The reason that's the narrative is because most orgs haven't thoroughly deployed AI for software development yet. And I'm really excited over the next year or two years as we deploy further into these larger organizations for people to see the actual impact that it has and to get this, like, I can already see there is going to be a burst of optimism that comes when people see what actually happens when you deploy agentic AI into these companies. And I think that's going to be really, really exciting. There is like way too much kind of like negativity or fear around this. And I'm just ready for everyone to be like, wait, this is like the most exciting time to be a developer.

Yeah. We're probably at the lowest number of developers there will ever be for the rest of human history.

Totally. Hot take.

I like it.

I love that.

Well, speaking of optimism, you're both based in San Francisco, building in person. We're recording live in San Francisco. What are you excited about personally as founders in the Bay Area?

Yeah, that's a good— that's a great question. Well, one thing is my girlfriend is going to match to be an OB-GYN.

Let's go.

She already matched.

Congrats. That's a huge week.

What? Well, so doesn't know yet, but on Friday she'll find out. Okay, so that's super exciting.

Vibes.

But probably like with respect to San Francisco, the city has evolved so much. Like my grandparents moved here in the '60s, left in the '70s. You know, they came straight from Mexico. And I think that the, the environment has obviously evolved so much. There was like a 3 to 5 year period where I feel like the whole country was trashing on San Francisco. And I think that we're on the up now. And so it's so great to see the city just like really, really, you know, we were here in the dark times.

It was dark.

We were in the same COVID bubble when no one else was on the streets. Yeah, it was interesting.

Yeah, yeah, yeah. I have been referred to and definitely self-identify as a San Francisco maximalist. I think it's the best city in the world. I think it is the only tier 1 city in the United States. And I think it is only going up from here. Like, you know, said it is by far the most beautiful city by far. Like, the most interesting technology is being developed here. And it drives me crazy when people hate on San Francisco. It is only getting better. And I'm excited to, you know, hopefully play a part in making it a better city.

I think we have some, you know, quote highlights for this episode.

The mayor probably wants to hear this.

The new one.

Yeah.

Let's, let's call Daniel. Well, as we come up on time, we're going to ask our hero question for the podcast. We ask that the answer is not a mom. We all love moms. But Matan, let's start with you. Who is a woman in your life that has had a profound impact on you and your career?

Yeah, this is a great question. So, um, certainly profound impact because this woman changed basically 12 years of my life. Um, this is Emmy Noether, uh, who is a physicist, uh, of fame from Noether's theorem, um, which basically states for every continuous symmetry there is a corresponding conserved quantity. It sounds a little jargony, but basically anything in the universe where you know, there's symmetry that is not discrete. So, you know, let's say there's— if we have empty space and we have, you know, a cube that's there, and then we start moving it forward or backward, the universe looks the same. That's a symmetry that corresponds to a conserved quantity, in this case linear momentum. Okay, so the description is not as beautiful. The mathematics behind Noether's theorem is so, so beautiful. Uh, it— like, I genuinely can't— if Anyone out there who is interested in math, go read Noether's theorem, like understand the math behind it, and you will probably maybe spend the next 12 years of your life pursuing physics because of how beautiful it is. And that's, you know, what eventually brought me here. And I'm very happy, you know, to be here. So shout out to Emmy Noether.

That's awesome.

You know, that's awesome. Yeah, this is, this is a, this is an interesting one. I think that probably the most profound impact there is She was working at OpenAI. I believe she's at Thinking Machines now. Lillian Wang has this blog, and it was probably like a week or two before we actually met. She published her blog post on prompt engineering, and then there was one like a month after on agents. And this is like the best technical blog that you could ever read. It's so dense, it's so filled with insights and alpha. And, and so reading that basically convinced me that I should go and work on LLMs agents like, like full-time. And that— and it was very timely. And I definitely credit that blog. Go check it out. It's awesome.

Yeah.

I just want to encourage you guys and I think all our listeners, I have heard a few times the serendipity of when you met and how the catalyst just came out of you to start your company. And I think frankly, Claudia and I had a very similar co-founding story. There was one week in July of 2020 where 3 different people mentioned to me that I should maybe start a podcast. And these were very different people. It was disparate sources. And Claudia was the third person. And it was a little voice note being like, I think we should maybe start a podcast.

Completely out of the blue.

Just totally random.

And I was like, oh my God, I think I have to start a—

we should start a podcast. And here we are 5 years later. And I just want to encourage you guys, like, so glad you took the leap. Clearly you're building so many amazing things. And for anyone else who's listening who's like, should I do the leap? Should I make the leap? I think we should all say that they should.

Yeah, well, I would say if you have to think about it, you shouldn't. And when it's one of those things, you won't even think about it. It's just like, this is what we're doing.

Yeah, absolutely.

If you know, you know.

Yeah, if you know, you know. Make the jump.

Yeah, 100%.

Well, you know, Matan, thank you so much for joining us today in person in The Room.

Thank you for having us.

Yeah, thank you.

Thank you so much for joining us at The Room Podcast. If you want more from The Room every week, subscribe to our newsletter at theroompodcast.com/newsletter.

We'll be back next week.

Week with a new episode and inspirational guest Tuesday. See you in the room. Perkins Coie supports the most innovative entrepreneurs and investors in fast-moving and high-growth sectors, addressing their myriad of legal needs. But the firm doesn't just provide end-to-end legal and business counseling to its startup clients. It also facilitates introductions to key advisors and sources of capital. Perkins Coie's interactive website, Startup Percolator, offers access to programs, resources, and rich dynamic content designed to assist entrepreneurs on their startup journey. To learn more, go to startuppercolator.com and perkinscoie.com.

This podcast is brought to you by Mercury, the banking platform businesses like The Room Podcast use to simplify their finances. It's time banking did more than hold your money. Now it can. With Mercury, you can pay bills in seconds, close the books faster, and even send invoices. Not only does Mercury do away with a patchwork of tools, it eliminates guesswork, giving you complete and accurate visibility into your business's finances, all from one account. Apply in minutes at mercury.com.