# Replit's President & Head of AI | Michele Catasta NEW ECONOMIES · 2026-06-16 Replit's President Michele Catasta explains how the company evolved from a 15-year-old open source side project to powering 85% of Fortune 500 by launching the first vibe coding agent in 2024, before the term existed. He describes the initial rough launch that attracted candid feedback, the scramble to build an enterprise sales team when IT departments knocked, and the belief that the coding problem is nearly solved—shifting focus to integrations. Catasta argues that technical moats are less important than execution moats built from a decade of battle scars and user trust. He predicts 2026 will be the year of 'agent managers,' where every role involves managing multiple AI agents rather than single-threaded work. The Visa partnership exemplifies Replit's strategy of providing reliable building blocks for monetization, allowing vibe coders to add payments with a single prompt. ## Questions this episode answers ### How did Replit become a leader in vibe coding before the term even existed? Michele Catasta explains that Replit launched its coding agent in September 2024—months before Andrej Karpathy coined 'vibe coding' in February 2025. The company had spent a decade building infrastructure and, despite a 'rough on the edges' first release, was first to market. This early move, combined with relentless shipping based on user feedback, positioned Replit as the category inventor and leader. [12:24](https://neweconomies.podhood.com/21c2b507-6735-4d23-b34f-6840e944841d?t=744000) ### What does Michele Catasta mean by 'agent managers' and why does he predict 2026 will be their year? Michele Catasta says 2026 will be the year of the 'agent manager,' where everyone—from ICs to executives—oversees multiple AI agents daily. The role shifts from doing tasks to launching, monitoring, and verifying agent output. This change, he argues, redistributes superpowers across the workforce and redefines productivity; those who don't adapt will be at a disadvantage. [26:50](https://neweconomies.podhood.com/21c2b507-6735-4d23-b34f-6840e944841d?t=1610000) ### Why did Replit have to scramble an enterprise sales team overnight? After Replit's agent became powerful enough for businesses in early 2025, Fortune 500 IT departments noticed employees building critical internal tools on the platform and demanded enterprise-grade governance. Without a sales team, Replit scrambled—hiring one salesperson and redirecting engineers to build compliance features—which led to 85% of Fortune 500 companies now using the platform. [23:18](https://neweconomies.podhood.com/21c2b507-6735-4d23-b34f-6840e944841d?t=1398000) ### What is the Replit and Visa partnership about? Michele Catasta describes a partnership where Replit integrated Visa payment rails as a building block into its platform. Users can now add a monetization feature to any app they build on Replit with a single prompt, gaining instant access to Visa's global, secure payment infrastructure. This collaboration aims to convert hobbyist projects into real businesses by simplifying the notoriously complex task of handling payments. [36:50](https://neweconomies.podhood.com/21c2b507-6735-4d23-b34f-6840e944841d?t=2210000) ## Key moments - **[0:00] AI Landscape** - [0:00] 85% of Fortune 500 companies use Replit, but many think the product is only 2–3 years old, reveals Michele Catasta. - [5:08] "I would love the public market to speak," says Michele Catasta on upcoming AI IPOs. - [6:44] Michele Catasta: Software creation is now possible without understanding code at all. - **[8:50] Replit Origins** - [9:00] Replit started as an open source side project by Amjad Masad in 2011, says Michele Catasta. - [11:35] Michele Catasta showed Amjad Masad a primitive AI agent demo that became Replit Agent three years later. - [12:38] Replit launched the first agent on the market in September 2024, months before 'vibe coding' was coined. - [13:06] Andrej Karpathy coined the term 'vibe coding' in February 2025, says Michele Catasta. - [14:45] "The first release was really rough on the edges. I was almost ashamed that we launched it," admits Michele Catasta. - [17:20] Michele Catasta proposed betting everything on building an agent, and Amjad Masad agreed, pivoting the company. - **[18:40] Vibe Coding** - [23:18] Fortune 500 IT departments knocked on Replit's door, forcing them to scramble an enterprise sales team overnight. - **[23:51] Enterprise Shift** - **[26:07] Agent Managers** - [26:50] "2026 is the year of agent managers," predicts Michele Catasta. - [30:36] "Every single person receives superpowers," says Michele Catasta, by using AI tools correctly. - **[30:53] Future Tools** - [33:28] Michele Catasta considers the coding problem almost solved, as models and agents become good enough. - **[33:42] Moats** - [34:11] "I don't believe in technical moats," says Michele Catasta. "It's an execution moat." - **[36:32] Visa Partnership** - [36:50] Replit partnered with Visa to let users add Visa payment rails to apps built on Replit with one prompt. - **[40:05] Inside Replit** - [41:40] Replit experienced a 5X growth in code generated in the last six months, says Michele Catasta. - **[46:27] Beyond Code** - [46:35] Michele Catasta is most excited about reliable self-driving cars and physical AI that understands the real world. ## Speakers - **Ollie Forsyth** (host) - **Michele Catasta** (guest) ## Topics AI Coding ## Mentioned Anthropic (company), Google (company), OpenAI (company), Rapid (company), SpaceX (company), Tesla (company), Visa (company), Waymo (company), ChatGPT (product), Claude Code (product), Cursor (product), GitHub (product) ## Transcript ### AI Landscape **Michele Catasta** [0:00] 85% of Fortune 500 are using Rapid, but a lot of people believe that Rapid has been around for 2, 3 years. It has been 10. Rapid started as a side project from Amjad Masad, and we were the first one on the market. We went through dark times before we did a bigger launch. The first release was really rough on the edges. I was almost ashamed that we launched it. But the moment you do it, people cared. They cared to give us very, very candid feedback. And we told each other, "Let's just make the product better, let's just target users, and see where things will take us." And that's what helped us, you know, to grow where we are today. 2026 is the year of agent managers. Everyone's role will change, where the average vibe coder can and is building advanced software tools. There is, like, a wide spectrum of what can be done, but what changes is that every single person receives superpowers. And, you know, these superpowers are expressed by means of you using AI tools correctly. If you don't do that, that's going to put you, you know, at a disadvantage from day one. And monetization is maybe far tricker than, you know, the average non-developer user believes. There are so many things that can go wrong. I believe that by the end of the year, everyone will have to. **Ollie Forsyth** [1:12] Michele, it's so awesome to have you here. I've been so excited to have you on the show for some time, and wow, so much has been happening in this AI space. But first of all, big welcome to NEW ECONOMIES: AI Show, and great to have you here. **Michele Catasta** [1:24] Thank you. Thanks for having me. **Ollie Forsyth** [1:26] So I didn't know about you, but the last, even 6 months feels like a complete blur on what's been happening across AI. Just in the last few weeks, we've had the likes of Anthropic vowed to go public, we've got SpaceX, who's also vowed to go public. So much is happening in driverless cars, trying to get everywhere and anywhere. The space for, and the category for, for legal space is heating up. You know, trying to get a dominant place. Just so much is happening. How far is that back? What is actually the state of AI today? **Michele Catasta** [2:01] I think what we're experiencing is literally how exponential curves behave, in the sense that when you have companies that grow 10x year over here, when you have technology that grows, you know, exponentially year over here, 6 months ago was literally ages ago, if you put that, like, on a linear growth. And that's why we're experiencing so much in such a short span of time. So I'm not too surprised that some of these milestone events that you mentioned are happening. If anything, it still feels that we're experiencing maybe too much in a short window of time. Like, it's almost hard to remember every single milestone that we've been seeing in the last few months. I think what really made the difference that, you know, not only people that are very much wired into tech, you know, but the rest of the world are experiencing, is the fact that this technology is actually starting to work extremely well. Like, especially when you package it in theright way, which is, you know, part of what we do for a living at Rapid, you know, it's our company mission. When you package it in a way that is easy to use and is, every user, you know, regardless of their technical background, can be very successful with it. That's when, you know, we open this Pandora's box of opportunity and, you know, we see it applied among every single domain. You know, as we were mentioning before, you know, it spans from legal to software development to, like, creating agents for your personal productivity. The list will keep growing over time. **Ollie Forsyth** [3:26] And it's all happening so quickly, as you said. This kind of whole Pandora box is really taking shape, and I think it's happening much faster than so many of us anticipated. But when the likes of OpenAI, who will eventually hopefully go public, and the likes of Anthropic, who will do the same, when they both go public, what does that actually mean for other AI companies who are maybe looking to do the same, or maybe just looking to catch up? **Michele Catasta** [3:55] I think OpenAI and Anthropic are in a very peculiar positionright now, in the sense that, you know, first of all, they're experiencing unprecedented growth, I would say, you know, in the entire history of tech. At the same time, their business model is very capital-intensive. They need to be able to keep building a lot of data centers, they need to acquire chips, and they need to have enough capital to train the next generation of models. So this is going to be a race that likely is not going to stop in the short term. You know, we're always going to want to have a higher level of intelligence packed into these very powerful, you know, frontier models. It's not the same for many other companies in this space. So what you call the application layer, where, you know, we stand to an extent as a company, and, you know, many other companies are blossoming in that layer, our requirements are not as capital-intensive. So it means that the dynamics of having to go public immediately, rather than remaining private, will be completely different for us compared to, you know, to Anthropic. That said, I'm excited it's happening because, you know, you keep hearing again and again that we're living in a bubble and that the investors are irrational and it doesn't make sense that AI is worth so much money. I would love the public market to speak. There is nothing better than having, like, a couple of massive IPOs and see how the public markets are reacting. And if the reaction is negative, perhaps it's about time that we all agree with each other that we're living in a bubble and we're going to go through an adjustment. But if that's not the case, it will really tell us that there is a general understanding that this is just the beginning of this technology. Like, what we're unlocking, what you saw in the last 6 months, it's really the beginning of how much the labor will be disrupted, in good and bad ways, by the way. You know, we can talk about it. But it's an event that I was waiting for, no matter what is going to be the outcome. **Ollie Forsyth** [5:50] It's so true. If the likes of Anthropic, OpenAI, go public. And it's interesting, Anthropic have kind of beaten OpenAI to it. They've kind of put their stamp on the approval and, you know, they beat them to it to go public. But that's how the game works. It's going to be really interesting to see, as you say, and the likes of Canva and, you know, others who are looking to go public. You know, this is really going to set the trend. But all of this is happening, you know, again, so quickly. What's really exciting to you at the moment in this whole AI space? **Michele Catasta** [6:21] Look, I'll be biased, of course, because I've been doing this for a living, literally, in the last 10 years of my life, you know. But I'm very passionate about AI applied to the world of software creation. And what we have experienced in the last, literally, 6 months is something I've been dreaming of for the past 9 years and a half. I knew we would have gotten here. No one can really predict the speed at which this would have happened, because it's also a byproduct of how much capital is being invested in these types of problems. But we reach a point, both for developers and non-technical users, where you can do quite a lot in terms of software creation without understanding code at all. Without understanding how to write it, without understanding how to maintain it. And of course, you know, I like to be very grounded to planet Earth. I'm not a person who's going to tell you, "With vibe coding, you can rebuild the next Google overnight with just 100 agents." We're not there yet. Who knows where we're going to get there? I love the present, because the present is what gives me the energy to wake up every single day and think about what kind of product we're going to be building. The present is already overwhelmingly exciting, because we have folks on Rapid every single day creating software that they always dreamed of, materializing business opportunity because of that, experiencing a level of creativity that they didn't have before that our product unlocks. So all this power is actually now being expressed by the product we have. And I know for a fact that the trajectory which these products are evolving and the models underlying are evolving is such that at the end of the year, the product that we have in mind is going to be far more powerful than what we launched a couple of months ago. So the present is almost, you know, completely focusing me and giving me, like, a narrow view. And it's great, I think, for builders, for operators. There hasn't been a great, most exciting time, you know, to create something. **Ollie Forsyth** [8:24] I agree. We were just chatting before we came on around, basically, each day feels like a blur,right? I mean, just 4 years ago, not even 4 years ago, ChatGPT launched, and ever since then, it just feels, it almost feels like the iPhone. Every 12 months there's a new release. But in this case, in the AI world, it feels like there's a new release every 6 weeks, or there's a, you know, massive milestone,right? **Michele Catasta** [8:47] It's literally 6 weeks. Yeah. **Ollie Forsyth** [8:49] Which is, like, just insane. Many people probably don't know this, but Rapid is actually 10 years old. I think most people probably think it's only a couple years old. What is the founding story of Rapid, and what is Rapid today? ### Replit Origins **Michele Catasta** [9:04] So Rapid started as an open source project, as a side project from Amjad Masad. I think the first commit that he had on GitHub is dating back around 2011. So we're talking about 15 years ago. And he maintained that project as, you know, as an open source repository called already Rapid back in those days, for 5 years. Until, I think, end of 2016, when, you know, he realized, you know, together with his today's wife, Haya, that there was potentially a lot of, like, business value. And he also realized this is such a powerful concept that I wanted to become my life mission. I want to build, and I want to bring the capability of creating software to everyone, not only, like, to the few people that love to be software developers in the world. So he took a leap of faith, incorporated the company, and at that point, you know, we started to live our lives on power trucks. I realized that years later, but myself, I just moved to the Bay Area in early 2017 to join Stanford. And there I was a researcher in AI, focused on software engineering. Back then, nothing was working. Like, I spent 4-plus years of my life sometimes, like, wondering my life choices. Like, why am I trying to, you know, push the boundaries of research in this direction if I can barely complete a line of code, if we can do things that are sort of trivial and developers don't get excited about? So I had to fundamentally believe on the exponential that I was mentioning before, the fact that eventually this technology would have reached a scale velocity and brought us where we are today. So the, you know, Amjad and I, the two of us, lived on these power trucks for roughly 5 years. He started to grow Rapid to be a very small team. It was a product mostly focused on edtech, so targeting schools, targeting students. They received a lot of goodwill. The community was growing in a very organic way. But as you know, schools don't have a lot of funding available. So, you know, none of them arrived to them and wrote 7-figure checks just to use Rapid, you know, in their school or in their district. But the exciting thing is that when we crossed paths, I was at Google. I was doing both research and running a team of engineers to productionize what eventually became LLMs. You know, they were not even called like that back in those days, almost. It was like the LLM team at Google was 50 people. I think at OpenAI it was 30. We were, like, a very small niche in the research world. We met, and I showed him a demo of what, 3 years later, became Rapid agent, a very primitive one, you know. But I built it on Rapid, and I told Amjad, "This is what I think we should bring to the world. This is so exciting. This is something that's going to allow to really drop completely the barrier of entry to software creation, and everyone will become a developer." Of course, we bonded because I went back to check the slide deck that he was using to raise money, even in 2016, and the thesis behind the company was exactly the same. It's the same mission we have today, empowering the next billion software creators. The underlying technological unlock would have been AI will generate the code, and everyone will just be focused on their creative process. So after we met, we became friends, I became an advisor of the company, and then I realized I have to be there. I have to go and build what we both believe, you know, will really change the course of the trajectory of the company. I joined, and then fast forward September 2024, we launched Rapid agent, the very first version, and we were the first one on the market. Like, back in the days, agents were hot on Twitter. People were posting about it. They were showing demos. But no one had a product that you could just use with a credit card. We did that. All of a sudden, we got the spotlight. We got a lot of people, you know, being excited about what we created. And I think, like, we charted and we explained to everyone what vibe coding would have become. Because vibe coding, as a term, has been coined in February 2025 by Andrej Karpathy. So we were, like, 6 months, you know, ahead of even the concept being born. And we created it, you know, we invented the concept, we made it popular, you know, across both technical and non-technical users. And the rest is history. Since then, the company's under the spotlight. That's why a lot of people believe that Rapid has been around for 2, 3 years. It has been 10. And the exciting thing is that before creating the agent, everyone at the company was creating all the infrastructure necessary to build what Rapid is today. So we were, like, getting ready for theright technological unlock to take us where we are today. **Ollie Forsyth** [13:44] I remember when Rapid was coming under the spotlight quite well. It was almost this whole flurry of, you know, memes or, you know, tweets were going around, and so many people were paying close attention. How or what were you thinking some of those, you know, moments that really put Rapid into the spotlight? Because it's fine having, like, a product, but you need, I feel like, so many different elements moving parts to actually get the wheel spinning, you know. **Michele Catasta** [14:16] You're totallyright. I think the spotlight, especially on social media lasted most a couple of days. Then people forget, even more today, because as we were mentioning before, there are too many daily news. Maybe back in 2024, it was slightly slower than we're experiencingright now. What was important for us to do, and the reason why I pushed the company to be the first on the market, is you start to receive a lot of feedback from your users. And fundamentally, especially at the beginning, the first release was not was really rough on the edges. I was almost ashamed that we launched it. But the moment you do it and the moment you are the only one out there, people cared. They cared to give us very, very candid feedback. And I love that, because I started to understand better what our product role should have been. I started to understand the assumptions that were wrong, the assumptions that wereright, and we were relentless. Every single day, we ship a new version, fixes, you know, improvements, and we went on this, you know, marathon. And it started that day, and it never ended. And when you get into this mindset that you don't just drop a product and everyone should love you, you're rather, like, running a long marathon, users realize that. And then they reward you, and then they keep giving you basically what you need. I'm not only talking about usage, you know. They're giving you the value. They're building the community around it. They're engaging. And that's what helped us, you know, to grow where we are today. **Ollie Forsyth** [15:47] And that's what you need,right? You need those super fans super early who basically help and build the company with you, which is, like, one of the most beautiful things. There was, like, this amazing I said it around this fintech called Monzo, which launched, like, 10 years ago. It was, like, pretty much exactly the same,right? **Michele Catasta** [16:05] We hired some of them, by the way. We hired some of our, like, very early adopters, people that were posting about us, doing videos. Now they're part of our community team. It's the least I could do to reward them for being so supportive from day one. **Ollie Forsyth** [16:18] Exactly. That's what you need,right? But I feel also it's so much about the team. And even though, I mean, we're nowhere near this because you guys,right? But part of my mindset is we have to push 2% every single day. Because not only have you compounded by 120% in 60 days, so much more is going to happen,right? Back in the early days, what was that team mindset, and how did you just keep pushing as they collected together? **Michele Catasta** [16:43] I would say we went through dark times before we did a big launch. Keep in mind that we were a late Series B company at that point, and we had an amazing amount of technology built. Like, the entire Rapid infrastructure is all owned by us. We have a tendency to build and find our own solutions rather than, you know, using a lot of vendors. So it was quite depressing to have that level of technical excellence without having found strong product-market fit. And I found, you know, the energy that the two co-founders of Rapid had back in those days to be extremely inspiring. Because, you know, rather than saying, "It's time to fold, it's time to return money, it's time to look for an acquisition," I talked to Amjad, and I said, "I think we should really bet everything behind creating an agent." And he said, "Let's do it." And we went. And then the way I kept, you know, the entire technical team motivated was, "We're going to be the first to do it, and it's going to be something that people have never seen." Even if we managed not to flip over the company and, you know, get on this amazing successful track we've been in the last couple of years, it's still something remarkable for your own, like, personal growth, for your resume. It's such a niche skill that today is extremely in demand in the market, you know. And if you're one of the people that built it at the beginning, of course, you know, you will be able to open every single door you want for your next jobs. So that's how I framed it, you know. You always want to have a master plan and a contingency plan. The master plan is, "We're going to take Rapid to the next level," which we did. But I think there is always a contingency plan if you do the best work in your life, regardless. And that's how I kept everyone motivated, including myself, you know. I have to be honest. Like, no one has certainty that when you launch something, it's going to get the level of excitement that we received. **Ollie Forsyth** [18:40] I feel the most incredible opportunity that the likes of Rapid and others have created in this vibe coding space is it doesn't matter who you are today. Anyone can now build a product. And this is, I think, being, you know, the best thing that has happened in the last few years. ### Vibe Coding **Ollie Forsyth** [18:57] But it does feel that the vibe coding, you know, genre era has, you know, changed a little bit. Where is vibe coding today, do you think? **Michele Catasta** [19:08] So when you started to hear about vibe coding very early 2025, I would say that the technology was enough to barely do prototyping. So talking about, like, landing pages, of course, you know, websites, but maybe, like, very simple product features. In the fast-forward 12 months, we live in a completely different world where the average vibe coder can and is building advanced software tools to the extent that, especially when we sell to, you know, businesses and enterprises, the vast majority of our users are creating internal tools. So the SaaS apocalypse that you heard about maybe 2, 3 months ago, you know, when the public market really had a big adjustment, you know, on some of those titles, on one end, I do believe it was an overreaction to, you know, how well certain AI companies were performing. At the same time, there is always, like, a hint of truth behind those big swings. And the truth is, a lot of the long-tail SaaS that, you know, software houses are building is being replaced by what vibe coders can create on products like Rapid. And it does make sense. Like, that software was always suboptimal in many ways, you know. It wasn't really custom-tailored to what a certain user wants. Maybe they had too many features. Oftentimes, it's overly expensive compared to what it offers. And our users love to have to build exactly what they need because it's not bloated. It has exactly the it maps basically exactly the business processes of their own company. And we see the success stories basically piling up day after day. So big leap in terms of what can be done in just a year. **Ollie Forsyth** [20:52] Exactlyright. And if we look at what's just happened just in the last 12 months, 2 years, imagine what's going to happen in the next 6 to 12 months,right? **Michele Catasta** [20:59] Of course. **Ollie Forsyth** [20:59] It feels like and it feels like there's been this pretty clear divide now where it's not just the hobbyist types of creators, builders who are building the products. It's now actually fully baked into the enterprise,right? What is the role of a product manager or developer engineer today? **Michele Catasta** [21:19] First of all, just to give an idea of the extent of utilization, 85% of Fortune 500 are using Rapid today. That's how widespread the product is within companies, 100%. And yes, PMs and designers tend to be the early adopters. I think it's part of also their persona. They are always eager to discover new technology. It's part of their work, you know. They need to understand where the world is headed. They need to try all the latest and greatest, you know, in the space. But what I'm getting really excited about is the fact that our aperture has become much wider. Now every ops team within our company is using Rapid. HR, marketing, sales, whoever has a need to replace part of their tedious work or to create the software that they can't really find on the market, they are hopping on Rapid and creating tools that are delivering a lot of value internally. So complete shift in terms of going from the early adopters of 2025 to, I would start to call, you know, like, mass adoption already today. **Ollie Forsyth** [22:28] The mass adoption, I think, has been I mean, that's crazy. So many Fortune 500 companies use Rapid in such a short period of time. **Michele Catasta** [22:40] And just a trivia, early 2025, you know, we're having a discussion with Amjad and some of the other execs, and we told each other, "It would be totally fine if we were a PLG company this year. Let's just make the product better. Let's just target users. Let's collect their feedback and see where things will take us." And the company started to knock at our door because, especially after we launched the second version of our agent in late February 2025, it had become powerful enough for a lot of users to build software that was really consequential for their business. And as you know, IT was starting to be concerned, you know, because they see a lot of things being built and deployed on Rapid. They don't have an enterprise account, so they wanted to make that happen. And guess what? We scrambled an enterprise team, you know, almost overnight. So we found one salesperson. We took out a few engineers and told them, "From today, you care about our enterprise roadmap." And, you know, we quickly built, you know, together all the table stakes features that you would expect. And from there until now, basically, enterprise has become one of our biggest, you know, focus. **Ollie Forsyth** [23:51] Wow. You know, we have a lot of founders who are building similar products and vibe coding AI tools, you know, which are for larger enterprises, but they focus on hobbyist creators initially. How do you think a good company or a good product transitions from a hobbyist type of customer profile into an enterprise type of profile? ### Enterprise Shift **Michele Catasta** [24:17] You'd be surprised, for us, the core product is pretty much the same. And I want to keep it that way for as long as possible. Like, I don't want to think of having a nerfed version of our agent for, you know, the hobbyist and then a more powerful one, you know, for the enterprise. We only have one setting, you know. We have different agent modes that are using different type of models and a different agent architecture. They come at different price points. But aside from that, even, like, the most powerful mode is available to a person who just, you know, onboarded on Rapid 24 hours ago. So my philosophy is the technology is so advanced and it has to keep making progress that it doesn't make sense to preclude part of our user base to, you know, to get the rest dirty with it. At the same time, being enterprise-ready doesn't mean only making the agent more powerful. It also means adding this constellation of features around it that make it, you know, viable for the enterprise to adopt. You know, I'm not talking about security, data governance, you know, the privacy, everything that matters a lot, you know, especially for big companies. So we've been working hard at that, you know, we're by far the most advanced vibe coding company when it comes to the level of compliance. But I don't want us to stop here because aside, you know, to make the business more solid and more viable, it is really exciting to sell to big companies. The type of feedback that they give and the fact that they really want to be successful after they buy one product is giving us a lot of valuable information on where to take Rapid next. And we interview them and we try to understand, "How do you use Rapid daily? What do you like? What do you dislike about it?" It's a gold mine of valuable information for us. ### Agent Managers **Ollie Forsyth** [26:07] And this is going to change for sure how product managers, engineers, everybody, you know, who's. **Michele Catasta** [26:13] Oh, especially. **Ollie Forsyth** [26:14] They're building products,right? **Michele Catasta** [26:16] Yeah. **Ollie Forsyth** [26:16] And it's definitely already changed. But then how their roles are perceived and shaped is going to change for sure,right? When I speak to my product and engineering friends who work at the large AI companies and large tech companies, just in the last 12 months, they've said, "This is what my role used to be. This is what my role is today." And they say, "Hey, Ollie, I think actually in the next 12 months, I'm not going to be the typical product manager I used to be. I think I'm just going to be managing and building my agents to make my role better." Is that accurate? And do you think that's where we're going next? **Michele Catasta** [26:50] Yeah, I think 2026 is the year of agent managers. That's how our everyone's role will change. Regardless if you are a manager in a company or an IC, the expectation is that everyone will have to manage several agents. Now, I went as far as saying in an interview that I hope the word manager sort of disappears, at least in the company jargon, because everyone is expected to manage resources. And yes, you might argue that, of course, managing humans comes with a different type of skill set compared to managing agents. But the mindset shift has to be the same. It's not about you as a person being sort of single-threaded and working on one big problem at a time. It's more about waking up in the morning, mapping exactly what you want to accomplish in the day, launching several agents, maybe building some of them specialized to what you're trying to accomplish, let them do the work. And then the rest of your workday is, like, is vetting what the agents are doing, having that critical thinking to understand if the results you're being put in front of an agent are actually correct or not. That is where a lot of our skills are going to be focused on in the future. **Ollie Forsyth** [28:04] And is that pretty much how your team operates now on a daily basis? **Michele Catasta** [28:09] 100%. Ranging across engineers to data scientists to people, sales, they're all in front of tools. We build, of course, our own tools. I think I spend maybe more than half of the time on my work laptop. I'm in front of Rapid tools. You know, aside, you know, from Gmail and Slack, you know, the standard productivity tools that are never going to be replaced, at least in the short term, everything else that we use at Rapid is built by us. And of course, you know, we have to partially because we want to look for our product and we believe profoundly in it. But I do see the difference. Like, even the way in which we're growing the company now is such that a lot of these tools allow me to have a to have middle management in a different way, not to have to grow the company so much in terms of hierarchy. And a lot of companies out there are changing exactly for the same reason. **Ollie Forsyth** [29:02] You know, I feel a lot of operators and people working in these larger companies, and they could even be small, smaller companies, still don't fully grasp the opportunity that lies ahead. For any operators who might be listening to this who might not be might not be playing with these AI tools, whether or not maybe they're slightly far behind, but even if they just have, like, the basic level of understanding and we think where the space is going next is around agent managers, how should they be thinking about their careers going forward, do you think? **Michele Catasta** [29:35] I believe if they are not investing a good amount of their week on learning how to use these tools and become more productive with them, fundamentally, they are hurting their career trajectory. And that's why, you know, whenever people ask me for recommendations, I don't even tell them use a specific tool or try to build X, Y, or Z. It's more get your hands dirty. Like, everyone is doing a different job, you know. Even if you're a PM, a PM at Google is completely different from being a PM at Rapid. But the nature of these jobs, regardless of where you are, is changing. And the exciting thing of being living in this period is that you, as an expert, can decide how your job is mutating over time. Like, you will be deciding within a company what kind of new PM you're going to be. Maybe you're going to be a person who interfaces more with external companies. Maybe you're going to be a person who helps us to have all the tooling to make projects run on time. Like, there is, like, a wide spectrum of what can be done. But what changes is that every single person receives superpowers. And, you know, these superpowers are expressed by means of you using AI tools correctly. If you don't do that, no matter how hard you work, you're not going to be able to work smart. And that's going to put, you know, as a disadvantage from day one. **Ollie Forsyth** [30:53] You know, what we can build today, like we were saying earlier, you know, you go to any of these vibe coding tools, they're pretty accurate now. I mean, just in the last 12 months, and I've used, you know, pretty much all of them, the accuracy is so good now. ### Future Tools **Ollie Forsyth** [31:05] The speed is so much better. What do vibe coding tools and the ecosystem tools that are surrounding this, where do they look like in the next 12 months? **Michele Catasta** [31:14] I wish I could tell you 12 months from today, but fundamentally, I believe that by the end of the year, a lot of the vibe coding tools will sort of not look all the same, you know, but the kind of expectations that users will have will apply across, you know, most of the top vibe coding tools out there. And the reason is we don't need the next leap in terms of model intelligence to make vibe coding 10 times better. Like, it's already great. Probably 90% of what an average Rapid vibe coder wants to do on our product is possible already today without much frustration. So say we arrive at the end of the year, we add another 9 to that number. So it's going to be 99% of the things we want to do as a vibe coder are going to be possible on the platform. We're kind of done in terms of pure AI power on what the agent can do. What really makes the difference is the entire ecosystem you build around that agent. And the reason is building software is not just about writing lines of code. More and more is about integrating services and, like, understanding how to have your software or your agent interact with the rest of the world. That's where the real complexity still lies on. And again, it's the same also for developers. Like, that's where they have to put most of their effort in understanding the architecture of their software. At Rapid, we strive on making also that part of the creation experience as easy as it gets. We have hundreds of connectors ranging from personal use all the way to data warehouse in the enterprise. We have a lot of partnerships, you know, with different companies out there that help us to make, you know, auth and logging easier and more reliable and payments more reliable across the world, you know. So we want to bring building blocks within Rapid so that rather than you having to think how to rebuild something from the ground up, you can easily integrate it into your application. And that's where a lot of the value of a company like us will be expressed in the months and years to follow, not just about writing code. I almost consider the problem solved, even though it kind of breaks my heart because, you know, I've been focusing on that for a decade. I think we're getting to a point where the models and the agents are starting to be really good enough. Maybe one year from today, most of my brain sagos will not be focused on that. They will be focused on the rest of the user experience. ### Moats **Ollie Forsyth** [33:44] You know, when these vibe coding tools first appeared on the market, I will be honest, I was like, these may make it. They're just these AI wrappers. How many of them are actually going to stay relevant? Okay, many of them are still staying relevant, but I think it's going to be really interesting to see the next 12 months. What is the biggest moat these vibe coding tools should be focusing on? And how does Replit create that moat, if so? **Michele Catasta** [34:11] I will tell you, I don't believe in technical moats. And this is not because fundamentally, I think Rapid doesn't have technical moats. So I'm coping with the argument. It's more about you could almost make the same argument about Frontier Labs today where, you know, they're extremely technically advanced. They have access to all the talent and the resources in the world. But if you put, you know, their models side by side, the differences are not mind-blowing. You know, yes, Anthropic has been at the top of the world when it comes to coding capabilities for practically two years. And now OpenAI is very close by. On some benchmarks, their latest model is also even better. Does it mean that Anthropic doesn't have any technical moat when it comes to training AI models? It really depends. I do think they have. I think, like, they have a lot of internal talent and a lot of skills to create, you know, something of amazing value. But, you know, the same argument could be applied to Rapid. Can another company provide enough resources and focus to rebuild Rapid? Of course, yes. You know, I think there are a lot of smart people around the world. We're not the only group of smart folks focused on this problem. But today we are. You know, that's all we do. And we have 10 years of history of knowing what works and what doesn't. We have a very large user base that tells us what to do and what not to do. We have a lot of battle scars, especially in the last two years. And that teaches you a lot also in terms of company culture, product philosophy, things that we should be adding in product versus things that we should be removing. That is our real moat. Like, understanding the space so well and having such amount of relentless focus and technical expertise that we've built, especially in the last couple of years, that it will be really hard to replicate anywhere else. And that's why, even though you've seen popping up tens of different vibe coding solutions in the last year or so, the first 80% is easy to clone. The rest, you know, the last mile, it gets really close to the customer. That's where the real complexity is. And we haven't given up. If anything, we go deep down in the rabbit hole and we keep building exactly what especially our enterprise customers want, that level of convenience and good integration is what makes the difference. And I consider that basically our moat. But it's an execution moat. It's not a technical moat. ### Visa Partnership **Ollie Forsyth** [36:32] Yeah, I kind of feel actually with these, you know, integrations and partnerships, that's potentially, you know, where a lot of these AI companies are going to go. You guys have just announced a great new partnership with Visa, which we should talk about. What was that all about? And that's awesome. **Michele Catasta** [36:50] Yeah, with Visa, it's a fun story. I love it because we did back first of all, we are in a very relatively small city in the Bay Area called Foster City. And Visa is historically the largest tech company there. And there are not many other tech companies. And we're becoming the second most visible tech company in the same small city. I can literally see their building here on myright, you know, outside of the window. So they started to use Rapid roughly a year ago. So when I told you, companies started to show some level of interest already, like, in Q1, 2025. And in the process of us visiting them and giving them some support to be more successful, you know, with the previous version of the product, we also started to talk at the partnership level and try to figure out, okay, what can we do together? Not only how can we make it more successful on Rapid, but how can you make us more successful as a product? And it turns out that, of course, one of our main goals is not only to allow people to create software. I think of software more as a means to a certain end. And the end that we all get most excited about is allowing people to create new businesses. That's what we want to empower people for, you know, create another source of income, maybe change in the way in which they're running a business that already exists. Of course, monetization plays a big role in that. Like, you create a product online, you want to make money with it. And monetization is maybe far tricker than, you know, the average non-developer user believes. You know, it's not just about taking my credit card number and moving money around. There are so many things that can go wrong. Security, making sure that, you know, the user exists and, you know, supporting payments across the entire globe. You can't think of a better partner than Visa for that. Like, they earned the user trust over several decades. And yes, in principle, as a vibe coder, you could literally write the code that connects you to the Visa APIs and actually create that monetization feature in our product. But as I was mentioning before, it is far more reliable and also more cost-efficient if we have building blocks. So we created one of these building blocks together with Visa. And today, you can take a payment solution, bring it into the application you build with Rapid, powered by Visa. We basically have one prompt and, you know, 100% reliability. You get all the benefits, you know, running on their own network. This is just step one. We have, like, an exciting technical roadmap ahead of us when it comes to our collaboration. But this is more like the type of approach that we have across, you know, different partnerships. We want to we think of Rapid as an ecosystem. It's a platform that runs one of the most powerful coding agents in the world. And unlike, like, a pure coding agent, let's say Claude Code or Cursor, it has a lot of different tools and integrations available so that if you want to accomplish a certain goal, we can be very opinionated on your behalf and tell you, you want to monetize your app, bring in a Visa payment rail into your application. And we're going to basically keep adding them on a regular basis in the next months. **Ollie Forsyth** [40:00] Super cool. Super cool. Excited to see how that, you know, kind of, you know, unfolds and also partnership, I'm sure, many more to come. You know, when I have all these amazing guests on the show, the favorite part that every listener loves to hear about is what actually goes on inside these companies. ### Inside Replit **Ollie Forsyth** [40:18] So if I was to join as your chief of staff for a day or for a week, which would love to do maybe at some point if I didn't have a job, but if I was to do that, what does a typical day and a week look like? **Michele Catasta** [40:32] So we start Mondays by going through all the company metrics and, you know, what is top of mind. We have a staff meeting where basically every team lead comes with the biggest issues that they're facing or biggest opportunities that they explored or, you know, biggest wins that we had in the past week. And we have this kind of, like, sync checkpoint, you know, to know where the company stands. And then we dive deeper into, like, different synchronization meetings, you know, based on the technology, on AI, on product. And we basically, like, dispatch tasks around, you know, the rest of the company. Then at Rapid, we have this concept of deep work days. So Tuesdays and Thursdays, we try to keep calendars as clean as possible, especially for ICs. Basically, for ICs, we make it mandatory. You have to feel guilty if you drop a calendar invite to any IC's calendar in those days. And it makes a difference because especially now, you know, we're talking about being manager of agents. Imagine you wake up as an IC, you start your swarm of different agents to create code. The botanic at this point has become reviewing the code, not generating it. That's how productive we are. You know, we have experienced, like, a 5X in terms of growth of code generated in the last six months. So unprecedented speed. Myself, you know, and a few other execs, of course, you know, spend, like, every single day trying to extinguish the different fires that, of course, are happening for a company of our size. And then the weekends on a Friday where instead of having our all-hands meeting, we call it weekly wins. And it's a moment to celebrate all the victories that we had during that week. So, you know, we start by talking about some thoughts regarding what's happening in the industry. Sound like the main highlights of the week. You know, like, when we announce the Visa partnership, we talk about that. But I can tell you very similar to what you experience on X today. Every single week at Rapid, you show up at that meeting. There's a very long list of exciting things that have happened to us. And, you know, every team lead goes on stage, talks about, you know, basically what the ship, what are the next big things coming. And then, you know, we wrap up the week, not always on a Friday. It depends, you know, what is the next deadline. You know, we are not working at a standard paceright now. Let's put it this way. There is a lot of passion and love for what we're building. So I don't even have to keep people motivated. You know, some folks are doing six days a week. Some folks are doing very long days. It's just there is so much to be done and no one wants to lose the opportunity of making us extremely successful. So the end, what you would experience as a chief of staff is mostly the energy that we have at the office. And you would have, like, to block and tag a lot of people that try to talk, you know, with me because everyone has something that they want to bring up to my attention, which is great. You know, it means that they care so much about what is going on and they want to solve problems immediately. It's kind of part of our company culture to really have a very high level of ownership. So that doesn't change at all, even though the company grew a lot, especially in the last six months. **Ollie Forsyth** [43:38] It's been insane to see how fast Rapid has grown just in the last six months. And if we look forward to the next 12 months, it's going to be probably a completely different story,right? But I feel even, like, for us, there is every day feels like a blur because so much is happening so quickly. How do you stay disciplined and entirely focused on a couple of product launches versus just multiple and hundreds? **Michele Catasta** [44:06] So if you go back and see our, you know, product history in the last couple of years, literally every six months, we have a major launch. And it didn't happen by chance. You know, maybe the first time it kind of happened because I picked a date on the calendar when I thought we would have done the first release of agent one. And it was a lucky date. It was Amjad's birthday. So, you know, we thought, okay, let's pick a lucky date and let's see if it's going to change the course of the company, which it actually did. From there, the reason why I have a major launch every six months is because it's a good balance between giving us maybe a couple of months to polish the initial release. Because you never launch something that is perfect. We do our best. You know, we have a very high quality bar. But despite all that, someone would find bugs. And as the product is used by more people, we'll find different ways to optimize it, to make it cheaper. So we go in this phase of continuous improvements. Like, several times a day, we ship a new version of the agent. From there, we keep part of our energies to keep working in that way, like, to keep polishing. But then I put us in front of a whiteboard. I usually write a vision document to what I believe we should be launching for the next big, you know, big product shift. And gradually, I arrange the entire company behind this new goal. And the reason why that timeline is meaningful is because if you try to do that every six months or so, sorry, every six weeks or so, fundamentally, you're just doing small iterations. You can't really take big swings and rethink from the ground up what the product does. If you wait a year, by then, you're completely irrelevant in the industry because there are so many new launches that even now, every time there is a small company doing something new internally and externally, our user asks me, when is Rapid going to launch exactly the same feature? So we always have to be on alert and figure out basically what is important for us to add. So six months proved to be, like, theright balance where we can keep things polished, a very high bar. And every now and then, rather than just improving things, we introduce in the market something that has never been seen. And I want us to keep working on that cadence. It is very important for us to innovate rather than just making things better. ### Beyond Code **Ollie Forsyth** [46:27] What excites you most outside of this vibe coding spaceship? **Michele Catasta** [46:35] Oh, so many things. I mean, I can't wait to see, like, self-driving actually working reliably across the entire industry. If you come here in the Bay Area, more and more cars are driving on the road. You know, we have Waymos, we have Teslas. I do see the benefit of being in a world where a lot of cars are not driven by humans anymore. And this comes from, like, I'm an Italian guy. I'm passionate about sports cars. It's such a nice hobby to have in life. But, you know, when you want to commute, you want safety. You want to, you know, you want to be feeling comfortable if you're tired rather than believing that you're risking your life. So that's a groundbreaking technology that I think we're not going to give up in order to bring it to literally everyone in the world. In the slightly longer term, very excited about the space of, like, physical AI. So creating models that have an understanding of the real world, not only powering robots, of course, that is the ultimate goal, but also models that don't understand exclusively, say, text, images, videos, but also, like, sceneries from the real world, understanding what kind of action should be taken, you know, by observing what humans and other objects are doing. There are a lot of exciting startups in that direction. And I think it's going to unlock a brand new type of products that we have never seen in the world. So maybe the only complaint that you can have today about AI is that the product factor almost looks the same. You know, we see a lot of AI bolted on products that already exist, like a better slide maker and a better word processor. But now imagine if you're going around with a device that understands everything around you, understands the discussions you're having, the location, the interactions we have with humans, and then, like, helps you on a day-to-day basis, you know, make theright decisions, don't forget to do certain things. It's kind of technology is going to have an impact across the fabric of society. And I know it's going to happen sooner than expected. So very exciting research direction. **Ollie Forsyth** [48:34] You know, on the driver's cars, I think this is probably one of the greatest categories to watch. What Waymo are doing and there are a bunch of, you know, more local players or, you know, just quite a few cropping up. I think this is going to be so interesting to see. You know, when I was in SF last and October last year, took my first Waymo and I showed my family back home in the UK. And they were like, what the hell is this? And you say, like, you know, from Italy, I mean, you Italians drive interestingly. Interestingly, shall we say,right? **Michele Catasta** [49:07] Yes. **Ollie Forsyth** [49:09] And, you know, I live out here in Bangkok and the driving here is, you know, super crazy. So I think these driver's cars and then potentially driver's trucks, I think that's going to be really interesting to watch as well. **Michele Catasta** [49:22] Yeah. No, 100%. Whenever someone is visiting the Bay Area, I always tell them, sure, go and check the Golden Gate Bridge, enjoy the nature, but you want to sit on a Waymo. Yeah. It's an eye-opening experience. **Ollie Forsyth** [49:37] It's so fascinating. Hey, Michele, this has been so cool. Loved hearing all about, you know, the Rapid story of where we're going in this crazy AI world next, all about the latest AI trends. And we can't wait to see what happens. And thanks so much for coming on the show. **Michele Catasta** [49:51] Thanks for inviting me again. It's been great. --- This library is powered by PodHood (https://podhood.com), the podcast website platform.