In this episode of Founded & Funded, Madrona Partner Vivek Ramaswami sits down with Jared Palmer — designer, developer, and founder of Turborepo (acquired by VercelHQ ), and former VP of AI at Vercel, which was a 2024 IA40 Winner.
Jared walks through his unique path from Goldman Sachs to Vercel, and how he combined finance, design, and engineering to create beloved developer tools like Formik, TSDX, and Turborepo, and v0.
The two dive deep into:
- Why vertical integration is the future of AI-native dev platforms
- The founding and launch of Vercel’s v0.dev
- How Vercel is positioning for a world with 700M code-generators, not just 28M developers
- What makes teams and products move fast
- Why “text-to-app” will soon become “text-to-business”
Whether you’re a founder building dev tools, a product leader thinking about AI-native apps, or a developer curious about the future of your craft — this episode is packed with lessons and foresight.
Listen on Spotify, Apple, and Amazon | Watch on YouTube.
This transcript was automatically generated and edited for clarity.
Vivek: So let’s get into this because I think there was an interesting thread we were just pulling on. We were both in the same summer analyst class in banking. I ended up sticking with it, and that’s a whole other thing. But you had an amazing evolution from starting in banking to building products, the developer world. Just take us through that transition. What were those early career moments for you?
Jared: My finance career started and ended at Cornell. I really enjoyed it. I think I did it for the wrong reasons. I think I was doing it for the… Well, I shouldn’t say that. I think I was doing it for the status of it. At the time, I grew up in New York City, and so nobody’s parents were software engineers. They’re all bankers and doctors and lawyers because that’s what people do in New York. And then when it came to college, all of what I thought were the super smart kids were in finance, and so that was the competitive thing to do. And me being a pretty competitive person, it just seemed like there, good thing to do and I was really good at it too. I did dabble with physics and math, but finance was like, okay, let’s actually get a job. And my brother works in finance too. He still does. He still works at Goldman, so it was like a family thing too, of course.
So I graduated at the top of my class at Cornell, and I get THE internship, which is Goldman Sachs, Investment Banking, Financial Institutions Group (FIG), although I think you actually did a little better…
Vivek: I don’t know about that.
Jared: Because you got TMT, which is what I wanted to do.
Vivek: In New York, FIG was the thing.
Jared: Well, yeah, so I wanted a TMT, but they’re like, well, FIG, which is very prestigious. It’s one of the most hardcore groups. It also was a very different group. As you know, in FIG, the valuation system, is very different. You don’t value banks off of the income statement. You value it off the balance sheet, and so you don’t get to use all the same models and systems. It’s totally black magic and totally isolated, and it is very pigeonholing, especially me. I was on the banks team, so I was doing investment banking for banks.
It was academically interesting. The hours were grueling, but it was a great experience. I learned how to work really, really hard. But I also came home one day from the internship and I remember this visceral conversation with my dad and I was like, “I don’t want to be anybody that I work for. They’re all on their third marriage.” All the partners, first of all, they like facts edits with… They mark them up. That was a sign. But they’re not going to their kids’ sporting events, they’re not going to the lacrosse game, they’re doing presentations, and they’re helping the world in some way. It wasn’t like a passion, “Oh my God, I need to go save the world.” It was just like I didn’t want to be these people. I also didn’t like the job, so I only lasted a summer.
Vivek: I get that. And it’s amazing how little has changed in many ways. But what’s interesting is, as you say, you started your career, very, very early part of your career in finance, but then you ended up deep in the developer role. Software engineering, but not just application, but deep on the developer side, Formik and Turborepo. What led you to that space? What was the evolution of your journey in the developer side?
Jared: So it came all from design first. But basically, there’s a gradual progression towards deeper and deeper into the stack. I was always the guy that did our frat t-shirts. My mom was the VP of design of Estee Lauder for many years in the ’80s and ’90s. My dad is a creative. He was a music producer at the time, now he’s a tech blogger, and he would do a bunch of commercial production as well, and they taught me Photoshop as a kid. And so I had this innate sort of design family, if you will. And so I was always the guy, yeah, do the frat t-shirts and do the invitations and stuff like that. So one of my fraternity brothers had this idea for an app and kind of like in Legally Blonde, like classic Elle Woods, I was like, “What? Like it’s hard.” Just crazy confidence. And so I just designed it in Photoshop, and that was the tool at the time, and it was fine.
Anyway, that led to a couple more apps and a couple more stuff. And basically coming out of college around 2013, I had some burgeoning freelance design opportunities in front of me, and I lived in Manhattan and I could live at home. I was very fortunate and privileged to be able to do that. And I was like, okay, I’ll go back to finance in a year. I’m just going to do this. And I have gotten some pretty sweet offers for some apps that I can design, and that was great. I did that. And I never went back to finance. But I kept designing, and I designed for my friends, designed for different applications, stuff like that. And my freelance portfolio grew and grew and grew.
And at the same time, there were some new prototyping tools that had come out. The company Framer, which is actually still around today, wasn’t initially its Figma-type site’s competitor. The original version of Framer was actually a prototype into. When I started using Framer, everything sort of just clicked. And I had taken some intro to programming courses at Cornell, but 101, 100 level, right? Nothing serious. But something about Framer just made it amazing because what it let me do is you could import your Photoshop layers and then animate them very quickly with a little bit of code, not like scary amounts of code, but a little bit of code. And you’d see it and it would feel like super high fidelity. You could hand it to a client and they’d be able to play with it before it was built.
So I was addicted to it.
Vivek: Mindblowing
Jared: Oh my god. And it’s funny because now V0 actually is kind of a similar interface with the preview on the right and the code, or now it’s AI in the chat on the left, but it’s still kind of the same thing in my mind, it’s just the AI version of it. But anyway, this grew and grew and grew, and eventually I started posting to the Facebook group of Framer and joining the community, going to the meetups and stuff like that. I was running the Framer New York meetup at the time, and this just grew into a freelancing job. And then I realized, about two or three years into it, I would make a lot more money if I just built the whole app.
And so I did. Just figure out how to do it. My prototypes were also getting so intricate and realistic that I was starting to use Firebase and other things like that. I remember getting a DM one day on Messenger or on Facebook at the time, and it was like, “Hey, I saw you posting in the Facebook group. My name’s Mikhail Lumens. I am a designer at Instagram. I would love for you to come in and interview to design at Instagram.” And I was like, “What?” Anyway, so I did. I didn’t get the job. I failed one of the interviews, but they also said that I just had the craziest resume ever. They were like, “We don’t know what to do with you.”
So anyway, moving along there, my freelance career kept going, and I kept going deeper into the stack, building full-stack applications. I teamed up with a couple of people in New York and sort of grew that into an agency for a couple of years. I was using my dad’s, some of my contacts there, too. And the way I would attract talent was to produce open source. I was seeing what Vercel and Facebook were doing and how they were doing it. They were able to attract talent by leveraging GitHub and using it to not only lower their HR and recruiting costs, but also share and influence the way that they want technology to be built. And I did the same thing.
And so that led me to, when I was working with our clients, whenever I would come up with some novel solution, I would open source it immediately, and that led to Formik and that led to, I don’t know, a bunch of other GitHub projects. Then, I also thought that GitHub was going to be my way to credentialize myself, because with my background, I’m a designer, finance guy, but no one cares if you’ve got a zillion GitHub stars and you’re the creator of blah, blah, blah. No one’s ever going to talk about my background ever again if my experience is like, “Oh wow, he designed blank,” or, “He’s the author of blah, blah, blah.” So I figured, okay, this is my ticket. And so yeah, I just got really deep into open source and again, also building a bunch of applications. And fast-forward five years, I like to say I feel like I built a lot of the internet because I just got so many reps in building so many different kinds of applications.
Vivek: I love that. I mean, I think it’s so unique. Like finance, design, and then open source. You’re combining so many different things. But we’re going to talk about V0, because that’s really exciting. But how did you even make your way into Vercel? You were sort of acquired into Vercel.
Jared: Yeah, that’s correct.
Vivek: What was that journey like? Did you know Guillermo before? Yeah, how did this happen?
Jared: It’s a funny story. So I was building a lot of applications at the time, and Next.js, which is Vercel’s flagship open source project. It is, I’m going to say the most popular web framework in the world. If you name the company, they probably used Next.js, Anthropic, Perplexity, Sora, the old ChatGPT, they’re all Next.js applications. But when I was working for my clients at the time, they were clients, not customers, Next.js was in it’s very much infancy, and I didn’t like all the decisions. And so I actually published my own competitor to it, Trolling Jerimo, I called it After. js.
Vivek: That’s one way to get noticed.
Jared: Definitely. I called it After.js and immediately got his attention, and that sort of started this long relationship that the two of us had. I didn’t actually like everything that Next.js was doing at the time. And so we built our own in-house, and we open-sourced it. And I believe that ours was actually decently popular. I think Coinbase was using our After.js or a while, and so was Microsoft Store for a minute. Well, but basically my most popular open-source project was called Formik, and it was a form library for React. It was the most popular form library for React, but I wanted to bootstrap it a little bit more. Fast-forward another year or so, and the pandemic happens. I got some term sheets, I turned them all down. Pandemic happens, and things were going okay, but not amazing. It turns out forms are really hard and it’s not a great business.
It’s an okay business. But while I was building it, I was running into a pretty difficult problem with build times. And this led to my next company, which is ultimately called Turborepo. So my build times were taking 10 minutes, and that was because I was deploying to AWS Fargate at the time, and there’s no way to roll a Fargate container in under 10 minutes. So if I changed my backend, it took 10 minutes to deploy. I’m very impatient. This was no bueno. I cannot do this. So it’s just me. I need to stay nimble.
So I hacked together some crazy CircleCI scripts, and it was some serious hacking. It was disgusting YAML file that was like thousands of lines long to figure out, okay, when I change this, only deploy that. And I figured there’s got to be some way that somebody solved this before. I mean, this is not a novel problem. It’d be sweet if I could only deploy the thing that’s actually changed. So I realized that yes, this has been solved before and it’s called a build system, and they’re very sophisticated. Google has one, it’s called Blaze internally. Bazel’s the open-source version of it. Facebook has one. They’re all stupid names by the way.
Vivek: I don’t know why that is. It’s weird.
Jared: They all have dumb names. Twitter’s is called Pants. That’s my favorite one by name. Facebook is called Buck. And they all, when I say this part, you’re going to be like, “duh, it obviously works that way.” When they make a change to Facebook, they don’t rebuild all of Facebook. They only build the part that’s different and the part that’s impacted by the change. Of course. But that’s not how the JavaScript ecosystem evolved.
And so this was a sort of Eureka moment for me. I was like, “Huh, well that’s interesting. I wonder if there’s something there.” And I looked and there was some stuff, but it didn’t have the right energy to it. It wasn’t the way I would do it. And so that’s what ended up becoming Turborepo. I was like, “Okay.” So I found a way to get the best of both worlds, which is this idea of let’s only do the least amount of work possible, be as lazy as we possibly can. Only incrementally do what’s necessary, and then cash as much as possible.
Vivek: So Turborepo takes off. Super important, highly efficient. How did this end up at Vercel?
Jared: So I did this all by myself, solo, and then I was going to raise a seed round, and Guillermo was like, “It’d be super sweet if you just joined Vercel.” And I was like, “Well, why don’t you pay me?” And then I actually never told this story. I also then took Guillermo’s offer to Netlify, which is Vercel’s biggest competitor at the time. And that started a bidding more. But I had a lot of leverage because I had term sheets, I had an acquisition offer, and then there was just like, okay, so at any point in either acquisition, I could just walk and just be like, “Okay, well, I’ll just take the round and go.” So ultimately, that went on all summer of 2021, and Vercel made an incredible offer that I just could not refuse, and I didn’t take a dime of funding and joined Vercel.
Vivek: That’s pretty amazing.
Jared: It’s pretty awesome.
Vivek: Does Guillermo know that story now?
Jared: Yeah, yeah, yeah. He knows that story now..
Vivek: He’s not going to be watching this podcast and saying, “Wait a second, I didn’t know that.”
Jared: No. But pro-tip… I’ll probably regret this with my future M&A. But pro-tip to any founder — a great time to sell your company is while you’re raising.
Vivek: Interesting. Why is that?
Jared: Because you have the most amount of leverage, right? Because you can always just walk away and say, “Peace, I’ll just see you in the next round. It’ll be more expensive next time.”
Vivek: Do you need a term sheet in place, or do you think you can just do it even if you’re fundraising?
Jared: I don’t know — it probably depends. The fundraise makes it quite real. If that makes sense?
Vivek: It does. Yeah. No, I think it’s funny. See some of your early M&A skills from 12 years ago, 14 years ago, coming in handy now, right? So now you’re in Vercel, you’re an ex-founder, you’ve been running your own business, set of businesses for a long time, and now you’re inside this fast-growing company, fast scaling company. What is that experience like going from someone who owned everything – to now you’re in a business.
Jared: Yeah, it was different. So initially at Vercel, I came in as a software engineer and-
Vivek: IC?
Jared: Yes, as an IC, and then I built out the Turborepo team over the course of… It wasn’t my first assignment, but I remember the transition away from coding every day was very strange for me. And the other thing that was very strange for me was being blocked. I never experienced this.
Vivek: Someone said no.
Jared: Someone said no to me. What do you mean I have to wait for something? I’m always used to, well, if I can’t solve it that way, I’ll just hit it again tomorrow or I will attack it in another direction or whatever. But when I got to Vercel, I was like, “Oh yeah, the person who’s responsible for Terraform is actually asleep right now because they’re in Europe, and you need to wait until tomorrow to talk to them.” And I was like, “Wait until tomorrow?”
Vivek: That’s crazy.
Jared: And Vercel’s one of the fastest-moving companies in the world, but we’re very distributed. And so time zones are a real thing. This was a very strange thing. I’m used to like, okay, I’ll go figure out, okay, this Terraform configuration and I’ll get it done. And then I realized that’s not an efficient use of time. So that was a big difference. I did adjust very quickly, but that was one that caught me off guard. The other thing that caught me off guard was I was always so sure I was going to get fired, and so I was super terrified of it. So I would always work extremely in public. And this was actually even different for Vercel, but it’s one of the things that my teams do, I’m very proud of. I just treated our public Slack channel like a group chat, and I didn’t really even care the entire company was seeing it, but this was radical at the time. This was like, “Who is this guy?”
Vivek: Radical transparency, right?
Jared: Yeah, exactly. Radical transparency. And I would commit war crimes on Slack.
Vivek: Right, right, right. So they’re like, you have to be here, right?
Jared: Every time I DM him, he’ll just say, “Let’s move to public channel.” I wanted receipts because I was so convinced that I was going to miss something or someone’s going to ask me to do something and I just going to miss it or not interpret it. So if it’s all in the open, well then I got receipts. There’s a paper trail.
Vivek: There’s a record. There’s a paper trail, there’s a record.
Jared: It became a superpower on my teams and it’s actually all of my teams work that way. Even the V0 team works that way right now. And it’s awesome. Because the thing that I did learn though is that you need a certain type of employee that will thrive in that type of team, and you need a, I call it a classroom environment if you will. Right. So there’s certain types of people that raise their hand in class and will get something wrong, and they’re okay with it. They’ll just brush it off and they’ll raise their hand again, and that’s totally fine. There’s other types of people where they raise their hand in class and that’s it, they get it wrong, they-
Vivek: It’s done.
Jared: They’re done for the day. It ruins their day. And there’s some people that will never raise their hand that know the answer. That’s not what we’re looking for. When people who raise their hand are okay with being wrong, can get over it, can check their ego at the door, and the more important thing is the mission, right? So that was what I ended up self-selecting my team for, if that makes sense?
Vivek: And even before getting into V0, you need leadership that allows you to do that. That makes it okay for you to raise your hand, get it wrong, you learn from that mistake and then you come back to the well.
Jared: Totally. But it ended up being actually a key thing that I think helps my teams work so fast. And I think that was the other thing that I learned that I was a little bit faster than others. Again, because I was completely self-sufficient, right? I had done my 10,000 hours, but it was on my own pace. This was different than other people’s pace.
Vivek: Right. So speed matters.
Jared: Speed matters.
Vivek: Especially in a place like Vercel. So okay, let’s take us to V0 — one of the fastest-growing AI apps out there. For folks that know, there’s plenty of people who are users of the product, they’re going to be listening to this podcast. What was the origins of V0? Did you have a team that you spun up that was thinking about a product like this? Take us through the beginnings and the roots of V0 and to where it is today.
Jared: I spent about a year building out Turborepo in Vercel, and then I went from IC to director of engineering for all of Vercel’s open source projects. Which that’s Next.js, that’s the React core teams, Felt, Turborepo, Turbopack and web tools. So the crown jewels included. I did that for about a year. And I was helping the Next.js team dog food, the latest version of Next.js. And instead of building a to-do app, I started playing with some AI stuff. And then I built what is now the AI SDK’s playground, and that led to the AI SDK, which is different than V0, but that came first.
And I knew the AI SDK was going to be big because it was a really big problem. Streaming was very hard. Working with LLMs is very difficult. And so the AI SDK came first. But then fast-forward to the summer… Anyway, after I launched the AI SDK, I basically got the green light to go with full in on AI and go take Vercel AI from zero to one. We didn’t know what that was going to be yet, but we knew that the AI SDK is probably a thing, and we bet big on this AI thing. It’s interesting because the Vercel was always the home of all the crypto apps, but we never really leaned into crypto. It just happened.
Vivek: You just made a home of that.
Jared: It just happened, of course.
Vivek: Yes.
Jared: But we’re not going to take this passive stance on AI. We’re going to go very hard on AI. And this was 2022 something like that. 2023 I guess.
But we had one rule, which was that no random acts of AI, no slop, it had to be pretty good. And we’d seen what other people had been doing, some docs, chatbots and stuff like that. And that was what we would call a random act of AI. And so there were a couple of ideas, there was basically two proposals that I made to Guillermo. One was called DevGPT, which was a ChatGPT slash Perplexity that was very much focused on just our audience – just developers. And it was going to have indexes of all the most popular docs that we would hand-index. And I would do retrieval, very much Perplexity. Whatever we could do for coding, best effort, talk to your Repo, stuff like that. That was in the summer of ’23. This is like GPT-3.5. It was pretty early. The other proposal was called Webjourney. And Webjourney was this idea was that, well, we were all in love with Midjourney, it’s amazing, right? Well what if you could do that instead of for images, you generate an image, what if you could generate user interfaces? What’s funny is we ended up doing both, but we ended up going with Webjourney. And we did some prototypes, and I soon realized that summer that a couple of things were key unlocks. The first was that these models are really good at HTML and they’re really good at Tailwind CSS, which is a special type of Inline CSS framework.
And Tailwind is really incredible because all of your style information, what it looks like in the user interface, is encoded and co-located right at each div that matters. That’s relevant. This is amazing for LLMs because you don’t need to separately stream a CSS file or something like that. And so I started rendering the output of GPT-3.5 Turbo and prompting it, only use Tailwind, only use HTML. And then we’re like, okay, well, if we can do HTML, we can probably do JSX, which is the markup language for React.
So after some prototypes, we did it. So we launched, we then couldn’t call it Webjourney, we launched V0 in 2023 and it was amazing. Just took off from there. But initially, it was this idea of going from text to UI, and the other reason why it was important, we call it SIUI, not even apps yet, is because we didn’t even do full code generation at the time. We just did markup. And that was an important constraint because it allowed us to render the user interface on the fly. And it allowed us to have that sort of Midjourney style pick one, still felt like some sort of image thing. And we launched that, and it really, really took off.
Vivek: And so at what point did it become full-stack text to application? Was that a journey after you had launched it?
Jared: Yeah, so we are always limited by the models. We’re always building a model generation ahead. And basically when we launched V0, I think we were using GPT-3.5 Turbo, GPT-4 320K, and we never actually got to GPT-4 Turbo because it wasn’t good at this problem space. I always thought GPT-4 Turbo was a smaller model than GPT 4 32K, but that’s just me. That’s what I think.
Vivek: The naming convention would make you think that.
Jared: Yeah, but still at a certain point though in context… The other big problem with V0, why we couldn’t do chat initially is because of context length. We only had 4,000 output tokens and 4,000 input or 8,000 input. And then GPT-4 16K and 32K came out. But right now we’re at a couple of hundred thousand, now some are even a million token context windows, you can actually fit something in there, but we couldn’t do that at the beginning, so we had to invent all these weird techniques to avoid that. So we actually couldn’t do chat. But fast-forward another nine months or so, and finally it was like, okay, now we can do chat. And I remember we rebased towards chat in the course of a month. We wrote the whole app.
Vivek: It’s fast for a big company.
Jared: Yeah, it’s pretty fast. Yeah, yeah. And it took us, I remember the stats it was nuts. It took us 10 months to get to our first million of ARR for V0, and then it took us 14 days to do the next and 14 days to do the next-
Vivek: Crazy.
Jared: … and 14 days to do that. It was nuts. And that was all after we launched chat.
Vivek: So you had pretty exponential growth after chat, obviously amazing growth even before then. That’s spawned a number of competitors, both ones that came out maybe a little bit before you guys, and then ones that have come out after. It’s a very crowded space at this point, but obviously there’s a reason for that. It’s a big market. And so how do you all think about V0 in this kind of competitive landscape in terms of the kind of folks that you’re going after, the market you’re going after, and then how it fits in the broader company?
Jared: So, backing up, we’ve gone from text to UI to now we’re at this text to app modality, and I think we’re going to get to text to business in the future.
Vivek: Watch out.
Jared: Wow. VC moment there.
Vivek: Easy. Let’s go.
Jared: So we’re text-to-app right now. And I think about it as there’s sort of two sides of this. You could argue or steal, man, it’s kind of Snapchat and Instagram stories and Snapchat exists. How do I say it’s not a problem? It’s not a problem, it’s an app. It exists and it’s a company and people use it. And that’s awesome. Instagram stories also exist and it’s also awesome. And we say it’s easy. It’s obviously not easy to build a high-quality photo-sharing application at internet scale. It is certainly not easy to build a high-performance text-to-app service, but at the same time, I actually think it is being commoditized. So what else are you going to bring to the table, if that makes sense? And for Vercel, it’s vertical integration. V0 will likely succeed because V0 is the vertical integration of coding framework, AI, editor and infrastructure. Super nuts. We own Next.js, we own V0, we just published this awesome post about how our models work. We design the website and the editor and we own the infrastructure of Vercel.
And so that is, iPhone like, if that makes sense? You think about vertical AI is a big deal, in the developer space, vertical AI means vertical around the framework and for us that’s Next.js. And so it’s the perfect thing to match our audience, if that makes sense?
And so as it relates to what our competition is doing, I think that there’ll be some that go towards more consumers, some that goes towards more developers and some that go in between. And I think they’ll explore all of them. Again, it’s a modality, so it’s going to be slapped in all different types of variations. But I’m not worried about too much competition because the market size is absolutely ginormous because we’re going from, I think I was looking at Perplexity the other day and it was like, how many total software engineers are there in the world? And the answer was like 28 million. I think that number from people who can, let’s define a software engineer in the future, is somebody who generates code. That number is going to the limit of Excel users, which is like 700 million people on Earth. So there’s a huge-
Vivek: The audience just gets so much wider.
Jared: Totally.
Vivek: And it’s like if you went from Vercel being thought of as a developer company, it’s still going to be a developer company. It’s just the audience for V0 is not just developers anymore. It’s everyone. It’s everybody.
Jared: It’s everybody. And this is what we have to figure out. And I do think there’s some decisions have to get made. I’ve always thought that you should always sell to your existing audience. It’s far easier. And we ourselves are the heaviest users of V0. It’s very funny, the pitch decks of our competitors for their rounds they were raising are like, “We’re going to be the next Vercel and we’re going to build Vercel on the back of this thing.” And well, we already have that, right? So it’s up to us to vertically integrate that.
And I think that’s where, ultimately, what’s super powerful up V0 is that you’re generating Next.js code. We develop Next.js. You’re in an editor that’s also Next.js app, which we’re using to make everything about it better. So you could do that as well. We’ve open sourced the AI SDK, which is the framework that we use to work with AI in V0. We’re about to release some products around our sandbox and virtual machines that you can use to build your own V0. And then when you click the deploy button in V0, you’re not getting some toy little app infrastructure; it’s the same infrastructure that it supported, like three Super Bowl ads this year.
Vivek: Nice. It’s seen scale.
Jared: That can burst scale to infinity. And that’s pretty cool. So I like our position in that sense in the developer experience world, but I also just think the market is so gigantic that everybody’s going to win.
Vivek: It’s an exciting time to be in this market for sure. Jared, let’s end with something that I’ve heard you’re very good at, which is spicy takes. And so two ways we can go with it. One, where are we in this AI hype cycle? I’m curious to get your take there, but also just generally what are things that you believe about AI or where we are in AI that you think others don’t believe or should believe?
Jared: I think that you’re going to have AI managers far sooner than people think. And let me explain why. So have you played with Codex?
Vivek: A little bit. Yeah.
Jared: So we have a system coming like this too. You enter in your task, and it goes and spins up an agent and it’s coding task in this case. It will download your repository, very similar to V0. You can launch multiple Codex tasks at one time. You can do the same thing for Devin and we’ll try to work. So play that out one model generation and say, okay, first assume it’s going to get good at that, which is a good and fair assumption because they’re going to do reinforcement learning. The model’s going to get better, smarter. And so let’s just presume that that is a way that you’re going to orchestrate some work.
So when you’ve got multiple of these tasks that get launched in parallel, which one should you do first? And so they’re, you’re going to have an AI engineering manager that works for you, if that makes sense? That’s going to be managing these little agents that are going to go off and do your tasks. If it gets sufficiently good at scheduling, I’ll call it that, right? And it gets sufficiently trained, it should actually be an amazing engineering manager in theory. In a couple of generations of model, it becomes incredible at planning. Have you ever driven in a Waymo?
Vivek: I’ve been in many, yeah.
Jared: You ever want to go back in Uber?
Vivek: No. Absolutely not.
Jared: Do you want to work for a human manager. They don’t respect you.
Vivek: This is the question we’ll end on — I think this is a very topical debate right now about what’s the role of the software engineer and the software developer in the future? And some folks are saying, I would not want my kids to get into computer science. And other people are saying, well actually we’re always going to have a developer. If you look out even three to five years, what’s your sense of what’s going to happen here in this space?
Jared: Man, I think that I would still learn to code. There’ll always be a market for people who get things done. If you can get things done, the more things you can get done, the better off you’re going to be. There’s always going to be a market for high agency people. And so that’s something you should hire for. That’s something you should foster and try to learn. And as someone who just sat there and Googled to learn to code, now it’s even easier. You should learn as much as you can about how to build amazing stuff and get things done. And you’re always going to be employed. And your job may look very different than the software engineer of today or a couple of years ago, we would go on Stack Overflow or Google something.
But believe me, when your AI manager tells you to go get some high level tasks done and then you’re operating these a swarm of managers, you’re going to need to know all kinds of stuff. And I think it’ll be very similar to… I’ll use this analogy, there are people who play guitar and there are people who could play Guitar… You even play Guitar Hero?
Vivek: Yes
Jared: They were more moving towards a little bit more Guitar Hero than guitar. That being said, so that’s one trend and the other trend is the way of writing. A lot of people write and a lot of people will generate software. They’ll also be professional authors, if that makes sense? And professional software engineers, I don’t think that’s going away, but a lot more people will generate code and you want to call them developers? Then yes, sure, fine.
They’ll have so much leverage. It’s incredible. That is the thing that I came back to is, like, yes, it’s not going end up being replaced, it’s just the ones that are great, just have 10X, 100X leverage now. Especially as things get more age agentic. And it seems like the other trends you’ll see is that seed rounds will get confusing. They’ll be much smaller and also much bigger. And the ambition of companies and the polish they’ll have, consumer expectations will skyrocket. So pricing and seeds may not be the greatest thing for SaaS, but you’ll also see extreme amounts of competition. Because again, you’re going to have a small team of people has never had more leverage in the history of the universe. And I think they’ll have as little leverage as they will have the rest of their lives today.
Vivek: Yes.
Jared: So that’s not going away.
Vivek: Totally. And every market, we’re seeing this for seed stage companies where every market has 15 competitors right off the bat.
Jared: Oh my gosh. It’s like V0 has-
Vivek: It’s like six years ago.
Jared: Every YC batch is like three different competitors.
Vivek: Well, hey, you are in a great place when all the companies out of YC and just every seed company is looking at Vercel and saying they’re doing something right. And it’s really incredible what you guys have built and so thank you very much for sharing your really unique journey into Vercel and what you’re doing today. Congrats on everything. Thanks so much for joining us.
Jared: Thank you.