RAISE Summit '26 | Michele Catasta, Replit

Clips
More from RAISE Summit '26

Michele Catasta

President & Head of AI

Replit

play_circle_outline Enterprise adoption of Replit for internal AI and coding workflows

play_circle_outline Unified Data Integration and Multi-Target Deployment: 400+ Connectors to Databricks, Fabric, S3, Google Cloud, Azure and Mobile Stores

play_circle_outline Replit Aims for $1B After 60 Million Users, Global Expansion in Europe, Brazil and Asia

play_circle_outline Replit Agents: Democratizing Agent Building with a Harness That Simplifies Technical Decisions for Non-Technical Users

Info

Michele Catasta, Replit

Michele Catasta

President & Head of AI Replit

search

>> Palo Alto Studio Connection, Silicon Valley and Wall Street. I'm John Furrier, co-host of The Cube here with Gabe Olave, my co-host. Welcome to the RAISE Summit coverage brought to you by Celadon with support from Tensodyne, D-MATRIX, and Argentum. Welcome back to the live stream here in Paris, France. I'm John Furrier, your host of The Cube. We're here for the RAISE Summit Day 2, winding down the last interview of the day, wrap the show up. No better than to have the leaders in coding, Replit's here, the president and head of AI for Replit, Michal Kastana. Good to have you on theCUBE. Did I get that right?

>> Thanks for having me.

>> Okay, close, close. So first of all, Replit is really one of the hottest stories here because AI infrastructure, which is dominating a lot of the conversations because the AI buildout's happening at rapid pace. Europe's a big part of it., but it's enabling more use cases, more users, more open source. Everything's coming together and more people are coding and creating. You guys have a leading product. I'm a customer. I love your product. Um, I love to hear that. Congratulations. Um, explain the momentum for people going to understand. You guys make it easy for people to code. That's the app we all know about, the coding. Talk about the momentum you guys have and how big it is and what the scope, the magnitude of the uptake.

>> Look, we recently crossed 60 million registered users on the platform. So just to give an idea of the magnitude, as you said. But what I'm excited about in terms of momentum is what's happening with the enterprise. So when we— early 2025, we thought it'd be great if we were just a PLG company. The product is growing and people are adopting it. As long as the product is better, that's all we need to do. And then it became powerful enough that enterprises started to knock on our door. And we realized why. It is really hard to make AI actually useful within the boundaries of a company. It is much more intricate than just having powerful frontier model tokens at your disposal. So we started to put a lot of effort in that direction. And as you know, Rapid.ai is a 10-year overnight success. We've been building infrastructure for 8 years. Leading to the moment that we actually launched RapidAgent in September 2024. And all that job work is starting to pay off now because now we know exactly how to deploy agents, you know, within the walls of an enterprise. Yeah.

>> And that work you did gets you scale, which makes a lot of sense, $60 million. But talk about the adoption in the enterprise because it's different. Every enterprise is different. They have databases. Sometimes they're not even really in knowledge databases and Data lakes are like state of the art and that's still not good enough. We're hearing more about graph databases, but there's a ton of data. Data feeds the intelligence. That is the key element.

>> Exactly. Yeah. Intelligence without data loses most of its value, as you said. We have been embracing this in the following areas. So you could either decide to build a brand new data governance layer on top of your application and then you would have to ask every enterprise to sort of replicate that effort. So they would have to take care of data governance on their hyperscaler. They would have to do exactly the same on Rapidly. Or what we're doing is we are allowing our customers to create Databricks applications as well as Microsoft Graphene apps that basically connect to their Fabric, you know, data lakes. And by doing that, we don't have to replicate the data governance efforts. Those apps are running on-premise. They inherit, you know, all the settings, all the different rules that a certain company follows. But we still make it very easy for everyone to create new software. So we really crush the barriers of what it means to create useful enterprise software. And all the bottlenecks that employees experienced before, they're now gone.

>> So they don't need to do a one-off just to integrate you in like in the old days with, hey, you know, go through a POC, AppSec review. All that is just now—

>> all that is gone.

>> So they're controlling all their governance and security and identity all on their side. Full control.

>> I wish I could show our dashboards, you know, from the day we enable access, you know, for a company to, you know, adoption started to grow exponentially. It's a matter of weeks, you know, back in the days, as you said, it's POC, several months, documents, regulation. It just, it just takes off on its own.

>> What about cloud and born in the cloud, AI-native companies that are using, say, AWS or Google Cloud? Can they take advantage of it too?

>> So today already we support several deployment targets. Google Cloud, Azure, Databricks applications. You also can build mobile apps, and we make it very easy for you to ship them both to the App Store and the Play Store. We're going to just keep adding new targets. It's a non-trivial amount of engineering, but it's an amount of engineering that we love to take over as a burden, because then we simplify life for our customers.

>> So I can just connect to an S3 bucket.

>> Of course.

>> And it's done.

>> Yeah, we have more than 400 connectors today. Some of them are for the consumer apps that you talk about all the time. Some others are, you know, different enterprise data warehouses, and we're going to keep expanding the list.

>> What's the coolest thing you've seen happen from a coding/connecting standpoint on the enterprise side? What are some cool examples that you can share?

>> Look, the stories I love, they mostly have to do with replacing SaaS solutions that those companies were buying back in the days. And of course, you know, one entry point is all about cost savings. You know, if you were buying a vendor for 7 figures and you can rebuild that on Rapid for a few thousand dollars, of course that's a massive success for the company that's using Rapid. The truth is, and this is also my own personal experience, when you can build a tool that is bespoke for how a company runs and really matches the culture, really matches their processes, is it becomes way easier to use. So I'm not a salesperson by any means. If you put me in front of Salesforce or an advanced CRM, I get lost. That's not my skills. No one loves it. It's like something necessary to have, but no one really can fall in love with a piece of software. If you install and you rebuild it in a way that, you know, is, is not bloated in terms of number of features, it uses the same jargon as the company, then even myself, when I use our own GTM tools at Rapid, I know how to use them. And how to navigate them correctly because they're part of the company culture at this point.

>> What is the common onboarding process? Take me through, uh, say in the enterprise for a second. I'm an enterprise, take me through. I'm interested, I love it. What do I do? How fast? What's the onboarding look like?

>> You will likely receive an invite to a hackathon. So we show up in person, we run a hackathon at your office. We usually ask for the biggest room we have available because they tend to overfill immediately. Folks from any different type of department are showing up, and in a matter of 2-3 hours, we let you go through the first "aha" moment. I know that you hack a lot and write software. I'm sure you remember the first time something you wanted to build actually worked. It's such an adrenaline rush, and we're basically allowing all of them to experience that. And then they start to connect the dots and they realize, "Oh, I always wanted to build this feature for our product. The engineering team never found the bandwidth to do so. Now I'm going to prove them this is actually useful.

>> I got to say, you know, one, one exciting consequence of your success has been you brought a lot of retired coders out of retirement because, you know, when you get older, you know, you learn coding, but you lose the syntax. What IDE do I use? Like, yeah, I'm running a company, I'm doing— I'm so busy, I got all these direct reports. Now I hear the story very common that they code prototypes of things they want to get done and then send it to the team and say, look, here's a draft.

>> Yeah, it can be done.

>> Demo, no memo.

>> Yeah, demo, no memo. Exactly.

>> It's like, okay, great. That has changed lives. I mean, it's been an amazing thrill.

>> A lot of our users are C-levels. We heard of a lot of meetings where the CEO shows up with a prototype. We have a mobile app as well. So just show it on the phone and say, I know this can be built. Go ahead and make it happen.

>> It's so easy. I did it. Exactly.

>> If I can do it, of course you can do it as well.

>> All right. Talk about some of the stats on the company. I'm curious. How big are you guys now? What's the expansion look like? Obviously, the success is well documented. 60 million users. Great enterprises. Adoption's up. Love connecting to the data lakes. Connect to the data. Just let the magic happen. What's some of the stats? Employees? Areas, obviously international. We're here in Paris. Take us through some of the data.

>> Recently crossed 300 people in the company. We've finally started to build a large go-to-market team, as you can imagine. So very excited to see new folks joining also from those teams. We, we target $1 billion revenue at the end of the year. Sounds very ambitious, you know, but we're on track to, to make that happen. And the expansion, yes, is happening globally. So that's one of the reasons, you know, why we are in Paris, why we are in Europe. I just came back from Brazil, Rio, just a month ago. It was a web summit. We have large communities that are growing also in Asia. So the product is by all means global and is also exciting to expand in global markets because you learn what different enterprise customers need.

>> Do you see geography and preferences change? And we hear that a lot with AI and certainly sovereign AI. There's different cultures, they have different orientations. And in Canva, I just interviewed Canva, uh, exec who's running EMEA for Canva. You know, Canva is a great product. Now they got a pro version, and he was telling me that every country has a different cultural vibe. The Germans behave differently than the Italians, you know. I know that. So they're— but they're— but it's just, it's a product localization.

>> Yes.

>> So yeah, do you see a lot of that in coding, or is coding universal?

>> Coding is universal to an But, you know, Rapid is far more than coding. It's a platform, it's an ecosystem. So we are starting to put a lot of effort on localization, and it's not just about the language, it's also about, you know, different ways in which we think about the UI. And also companies, you know, have different requirements depending on the geo. And so we are— we have like a very ambitious enterprise engineering roadmap, you know, all the different features that they need us to support on the platform side.

>> I love the transition because once you're a platform The world gets better because you're doing a lot more. So as you guys grow, you put— what are some of the features you guys are adding into the platform to benefit the users and the enterprise?

>> So I would say what I'm most excited about is the fact that we support all these different deployment targets at this moment. But overall, the, you know, our real reason to exist is the fact that we bake in governance in AI by being a platform. So there are a lot of very powerful coding agents out there, and some of them are definitely tailored towards software engineers rather than knowledge workers as we do. They give you a lot of flexibility and a lot of power, which is great for developers, but I will argue those are the only folks that really understand, you know, what it means, when it makes sense to deploy, how to deploy, what are the, you know, kind of data sources that they could be using. Everyone else at the company doesn't really grasp that. So if you put them in front of a firewall, so like very high intelligence tokens, but they don't really know what are the guardrails that they should follow, that's where you completely lose control of your infrastructure. And that's when companies panic because, you know, they—

>> Yeah, because someone ships something in the production.

>> Yeah.

>> And then you imagine when they unwind it. So you got to make sure there's protection.

>> Exactly. So Replit comes with a level of governance baked into the product. So say you are one of enterprise customers, we give you access to an admin dashboard. You have the level of granularity of deciding these are the only connectors that are allowed to use, these are the scopes for the connectors, this is the only deployment target that you can use, maybe all your deployments must be private. So that's the level of governance that companies need today in order to embrace AI at the level of coding. Coding is infinitely powerful, but it's also infinitely dangerous if you don't put the right guardrails around it.

>> And the benefits if you do it right, it's the creativity, it's the invention.

>> Of course. At that point, you stop thinking about everything else that is tedious for software creation and you really focus on what's fun, the creative process.

>> I like the positioning. It's very cool. On the connector side, you mentioned you got a lot of different connectors. What are some of the popular ones and what's coming? What are you working on?

>> So I would say the most popular ones are the data warehouses, because as you said, if you feed intelligence with data, that's the perfect recipe. Then, then you really unlock enterprise value. So every data warehouse that comes to your mind, it's already there. New connectors, I would say there's a lot of interest in new platforms and especially like new agents. But what I'm excited about is this large number of agent-to-agent protocols that are popping up in the industry. I wish there was a bit more alignment across companies because I probably count 4 of them at the moment and I don't know which one to embrace, but I can see a future in a few months where most of our agent usage will be driven by agents, not by humans.

>> Yeah, and the harness is a big topic. Harness selection could be a big deal. Yeah, especially in the enterprise. Cost per tokens. I mean, cost for the models. I mean, I just saw a study this morning from the CEO of Databricks, Ali Ghodsi, and he— they did their own internal benchmark on all their source code and they found that the models are pretty much the same in price. It was harness selection that drove it. And that was interesting. I didn't see that coming. That points to some of the nuances around backend. And so that works, then the frontend platform enablement happens. And so that's a good thing. What about harnesses? You guys take care of that for users, or is that on the enterprise side?

>> We bring our own harness. You know, we've been putting more than 2 years of research on that, and we, we own the harness for 2 reasons. First of all, we build it for non-technical users, so our harness is more opinionated, more autonomous, and really does its best not to put you in front of technical decisions, because those are the decisions that become blockers, and that's when you stop using the product.

>> Yeah, yeah. Which MCP server should I use?

>> Exactly. Imagine someone in marketing in front of the question, it would just close the tab and disappear. But the other reason why it's also very important to own it is because when you want to build an harness that is capable of interacting with the entire platform, then the, you know, the outreach is much bigger than a pure coding agent. So I don't think of our agent anymore as a coding agent. It's literally like an agent that is capable of using the entire platform, you know, correctly.

>> Yeah, it's a building agent. You build, you build products.

>> Yeah, it's exactly the entire software development life cycle. Cycle, not just creating PRs for a certain repository.

>> All right, so what's your goals now? You got the growth strategy, got a billion dollars you're eyeing for the revenue. What are some of the goals that you're working on? What's your focus? What are you optimizing for?

>> I'm focusing on agents, and what I mean by that is not the Rapid Agent. I want everyone to be capable of creating their own agents. If you think about it, it's such a massive unlock to be able to create an agent. It's literally what drove our success, and I want to bring that capability for everyone.

>> Everybody wants an agent. Who doesn't want an agent to work for them?

>> I build them for my own work. I'm already overwhelmed with information. If I didn't have them, it would be even worse.

>> Check all my email for me, tell me what I gotta do.

>> Delegate tasks at the right time. But agents are really hard to build correctly, especially with all the governance requirements we talk about and reliability. And there are very few companies in the world that have experience to do it in depth, and of course we are one of them. Our take, of course, is always very aligned with our company mission. This is— this should be a technology that is in the hands of everyone. So the next big launch we're going to be talking about, you know, maybe in a few months, invite me back, you know, we can do it.

>> I'd love to.

>> Yes. Yeah, same revolution we brought to app building is going to come very soon to agent building.

>> Are you going to have that preview on the consumer side first, or the individual side, or enterprise at at the same time? Because individuals might not have a governance issue, or small business might not have a governance issue. And this is where I think it shines, because if I have everything in Amazon, I got one governance policy already set. Is that going to be the first place, or are you going to do it all together?

>> I've been thinking about it exactly in that way. It's always easier to ship things in front of consumers and then learn the lessons there and then bring it to the enterprise. The difference between when we launched a couple of years ago is the fact that Enterprises now have visibility on what we're doing and they're so eager to get new features. So I'll do my best to make those launches almost overlap.

>> Yeah. I mean, you have to do the enterprise. One is great market into what you're doing, democratizing the creation. But also there's a lot of shadow AI going on.

>> Yeah.

>> You know, like it's like shadow IT with cloud. We saw that before. There's tons of it. The finance teams vibe coding and coding up stuff. Stuff can break. You got to rein that in.

>> Yeah, yeah, yeah.

>> Okay, final question. RACE Summit, obviously you guys were featured. I saw on the schedule AI infrastructure, a wide-ranging AI conversation up and down the stack. You guys obviously at the front lines of that on the frontier side. What's your take of the show this year? What's your opinion? What's your summary? If you were a reporter, how would you report this show?

>> First time for me at the Race Summit. I really enjoy the fact that it's full stack, as you said, you know, you get lost across the rooms and you see all the way from the GPU buildouts to the app layer like us. So great to meet founders across the entire stack. I'm European. I'm very proud to see like a flagship AI event happening in Europe. Paris is an amazing city. It was also fun for me to be like on a flight from SFO to here and meet Alfred DeFalksen on the flight, one of my friends in San Francisco, you know, like all founders in my space. It's great to meet you.

>> The whole tribe was coming from the Bay Area. You were at business class, all the business class seats were taken.

>> Yeah, but I think all of us are here especially to meet companies that are not exposed to what's happening in the Valley, but they're very eager to understand and embrace this level of innovation as soon as possible.

>> Yeah, and the show had a great success, the event on the opening night was pretty impressive. Yeah, Palace of Versailles was great. Fireworks, great stuff. Well, congratulations. We'll definitely want to follow up on your news. Looking forward to it. Thanks for coming on and wrapping up our show.

>> I'll be back.

>> All right.

>> It's an honor.

>> All right. We're wrapping up. This is a wrap for Ray's Summit Day 2 closing. I'm John for your host, Cube. Thanks to the team, guys. Great job out there. You got Gemma Allen, Alicia Volante, all doing a great job. Liam, all doing great work. See you next time.

>> Good.

Michele Catasta, Replit

search

>> Palo Alto Studio Connection, Silicon Valley and Wall Street. I'm John Furrier, co-host of The Cube here with Gabe Olave, my co-host. Welcome to the RAISE Summit coverage brought to you by Celadon with support from Tensodyne, D-MATRIX, and Argentum. Welcome back to the live stream here in Paris, France. I'm John Furrier, your host of The Cube. We're here for the RAISE Summit Day 2, winding down the last interview of the day, wrap the show up. No better than to have the leaders in coding, Replit's here, the president and head of AI for Replit, Michal Kastana. Good to have you on theCUBE. Did I get that right?

>> Thanks for having me.

>> Okay, close, close. So first of all, Replit is really one of the hottest stories here because AI infrastructure, which is dominating a lot of the conversations because the AI buildout's happening at rapid pace. Europe's a big part of it., but it's enabling more use cases, more users, more open source. Everything's coming together and more people are coding and creating. You guys have a leading product. I'm a customer. I love your product. Um, I love to hear that. Congratulations. Um, explain the momentum for people going to understand. You guys make it easy for people to code. That's the app we all know about, the coding. Talk about the momentum you guys have and how big it is and what the scope, the magnitude of the uptake.

>> Look, we recently crossed 60 million registered users on the platform. So just to give an idea of the magnitude, as you said. But what I'm excited about in terms of momentum is what's happening with the enterprise. So when we— early 2025, we thought it'd be great if we were just a PLG company. The product is growing and people are adopting it. As long as the product is better, that's all we need to do. And then it became powerful enough that enterprises started to knock on our door. And we realized why. It is really hard to make AI actually useful within the boundaries of a company. It is much more intricate than just having powerful frontier model tokens at your disposal. So we started to put a lot of effort in that direction. And as you know, Rapid.ai is a 10-year overnight success. We've been building infrastructure for 8 years. Leading to the moment that we actually launched RapidAgent in September 2024. And all that job work is starting to pay off now because now we know exactly how to deploy agents, you know, within the walls of an enterprise. Yeah.

>> And that work you did gets you scale, which makes a lot of sense, $60 million. But talk about the adoption in the enterprise because it's different. Every enterprise is different. They have databases. Sometimes they're not even really in knowledge databases and Data lakes are like state of the art and that's still not good enough. We're hearing more about graph databases, but there's a ton of data. Data feeds the intelligence. That is the key element.

>> Exactly. Yeah. Intelligence without data loses most of its value, as you said. We have been embracing this in the following areas. So you could either decide to build a brand new data governance layer on top of your application and then you would have to ask every enterprise to sort of replicate that effort. So they would have to take care of data governance on their hyperscaler. They would have to do exactly the same on Rapidly. Or what we're doing is we are allowing our customers to create Databricks applications as well as Microsoft Graphene apps that basically connect to their Fabric, you know, data lakes. And by doing that, we don't have to replicate the data governance efforts. Those apps are running on-premise. They inherit, you know, all the settings, all the different rules that a certain company follows. But we still make it very easy for everyone to create new software. So we really crush the barriers of what it means to create useful enterprise software. And all the bottlenecks that employees experienced before, they're now gone.

>> So they don't need to do a one-off just to integrate you in like in the old days with, hey, you know, go through a POC, AppSec review. All that is just now—

>> all that is gone.

>> So they're controlling all their governance and security and identity all on their side. Full control.

>> I wish I could show our dashboards, you know, from the day we enable access, you know, for a company to, you know, adoption started to grow exponentially. It's a matter of weeks, you know, back in the days, as you said, it's POC, several months, documents, regulation. It just, it just takes off on its own.

>> What about cloud and born in the cloud, AI-native companies that are using, say, AWS or Google Cloud? Can they take advantage of it too?

>> So today already we support several deployment targets. Google Cloud, Azure, Databricks applications. You also can build mobile apps, and we make it very easy for you to ship them both to the App Store and the Play Store. We're going to just keep adding new targets. It's a non-trivial amount of engineering, but it's an amount of engineering that we love to take over as a burden, because then we simplify life for our customers.

>> So I can just connect to an S3 bucket.

>> Of course.

>> And it's done.

>> Yeah, we have more than 400 connectors today. Some of them are for the consumer apps that you talk about all the time. Some others are, you know, different enterprise data warehouses, and we're going to keep expanding the list.

>> What's the coolest thing you've seen happen from a coding/connecting standpoint on the enterprise side? What are some cool examples that you can share?

>> Look, the stories I love, they mostly have to do with replacing SaaS solutions that those companies were buying back in the days. And of course, you know, one entry point is all about cost savings. You know, if you were buying a vendor for 7 figures and you can rebuild that on Rapid for a few thousand dollars, of course that's a massive success for the company that's using Rapid. The truth is, and this is also my own personal experience, when you can build a tool that is bespoke for how a company runs and really matches the culture, really matches their processes, is it becomes way easier to use. So I'm not a salesperson by any means. If you put me in front of Salesforce or an advanced CRM, I get lost. That's not my skills. No one loves it. It's like something necessary to have, but no one really can fall in love with a piece of software. If you install and you rebuild it in a way that, you know, is, is not bloated in terms of number of features, it uses the same jargon as the company, then even myself, when I use our own GTM tools at Rapid, I know how to use them. And how to navigate them correctly because they're part of the company culture at this point.

>> What is the common onboarding process? Take me through, uh, say in the enterprise for a second. I'm an enterprise, take me through. I'm interested, I love it. What do I do? How fast? What's the onboarding look like?

>> You will likely receive an invite to a hackathon. So we show up in person, we run a hackathon at your office. We usually ask for the biggest room we have available because they tend to overfill immediately. Folks from any different type of department are showing up, and in a matter of 2-3 hours, we let you go through the first "aha" moment. I know that you hack a lot and write software. I'm sure you remember the first time something you wanted to build actually worked. It's such an adrenaline rush, and we're basically allowing all of them to experience that. And then they start to connect the dots and they realize, "Oh, I always wanted to build this feature for our product. The engineering team never found the bandwidth to do so. Now I'm going to prove them this is actually useful.

>> I got to say, you know, one, one exciting consequence of your success has been you brought a lot of retired coders out of retirement because, you know, when you get older, you know, you learn coding, but you lose the syntax. What IDE do I use? Like, yeah, I'm running a company, I'm doing— I'm so busy, I got all these direct reports. Now I hear the story very common that they code prototypes of things they want to get done and then send it to the team and say, look, here's a draft.

>> Yeah, it can be done.

>> Demo, no memo.

>> Yeah, demo, no memo. Exactly.

>> It's like, okay, great. That has changed lives. I mean, it's been an amazing thrill.

>> A lot of our users are C-levels. We heard of a lot of meetings where the CEO shows up with a prototype. We have a mobile app as well. So just show it on the phone and say, I know this can be built. Go ahead and make it happen.

>> It's so easy. I did it. Exactly.

>> If I can do it, of course you can do it as well.

>> All right. Talk about some of the stats on the company. I'm curious. How big are you guys now? What's the expansion look like? Obviously, the success is well documented. 60 million users. Great enterprises. Adoption's up. Love connecting to the data lakes. Connect to the data. Just let the magic happen. What's some of the stats? Employees? Areas, obviously international. We're here in Paris. Take us through some of the data.

>> Recently crossed 300 people in the company. We've finally started to build a large go-to-market team, as you can imagine. So very excited to see new folks joining also from those teams. We, we target $1 billion revenue at the end of the year. Sounds very ambitious, you know, but we're on track to, to make that happen. And the expansion, yes, is happening globally. So that's one of the reasons, you know, why we are in Paris, why we are in Europe. I just came back from Brazil, Rio, just a month ago. It was a web summit. We have large communities that are growing also in Asia. So the product is by all means global and is also exciting to expand in global markets because you learn what different enterprise customers need.

>> Do you see geography and preferences change? And we hear that a lot with AI and certainly sovereign AI. There's different cultures, they have different orientations. And in Canva, I just interviewed Canva, uh, exec who's running EMEA for Canva. You know, Canva is a great product. Now they got a pro version, and he was telling me that every country has a different cultural vibe. The Germans behave differently than the Italians, you know. I know that. So they're— but they're— but it's just, it's a product localization.

>> Yes.

>> So yeah, do you see a lot of that in coding, or is coding universal?

>> Coding is universal to an But, you know, Rapid is far more than coding. It's a platform, it's an ecosystem. So we are starting to put a lot of effort on localization, and it's not just about the language, it's also about, you know, different ways in which we think about the UI. And also companies, you know, have different requirements depending on the geo. And so we are— we have like a very ambitious enterprise engineering roadmap, you know, all the different features that they need us to support on the platform side.

>> I love the transition because once you're a platform The world gets better because you're doing a lot more. So as you guys grow, you put— what are some of the features you guys are adding into the platform to benefit the users and the enterprise?

>> So I would say what I'm most excited about is the fact that we support all these different deployment targets at this moment. But overall, the, you know, our real reason to exist is the fact that we bake in governance in AI by being a platform. So there are a lot of very powerful coding agents out there, and some of them are definitely tailored towards software engineers rather than knowledge workers as we do. They give you a lot of flexibility and a lot of power, which is great for developers, but I will argue those are the only folks that really understand, you know, what it means, when it makes sense to deploy, how to deploy, what are the, you know, kind of data sources that they could be using. Everyone else at the company doesn't really grasp that. So if you put them in front of a firewall, so like very high intelligence tokens, but they don't really know what are the guardrails that they should follow, that's where you completely lose control of your infrastructure. And that's when companies panic because, you know, they—

>> Yeah, because someone ships something in the production.

>> Yeah.

>> And then you imagine when they unwind it. So you got to make sure there's protection.

>> Exactly. So Replit comes with a level of governance baked into the product. So say you are one of enterprise customers, we give you access to an admin dashboard. You have the level of granularity of deciding these are the only connectors that are allowed to use, these are the scopes for the connectors, this is the only deployment target that you can use, maybe all your deployments must be private. So that's the level of governance that companies need today in order to embrace AI at the level of coding. Coding is infinitely powerful, but it's also infinitely dangerous if you don't put the right guardrails around it.

>> And the benefits if you do it right, it's the creativity, it's the invention.

>> Of course. At that point, you stop thinking about everything else that is tedious for software creation and you really focus on what's fun, the creative process.

>> I like the positioning. It's very cool. On the connector side, you mentioned you got a lot of different connectors. What are some of the popular ones and what's coming? What are you working on?

>> So I would say the most popular ones are the data warehouses, because as you said, if you feed intelligence with data, that's the perfect recipe. Then, then you really unlock enterprise value. So every data warehouse that comes to your mind, it's already there. New connectors, I would say there's a lot of interest in new platforms and especially like new agents. But what I'm excited about is this large number of agent-to-agent protocols that are popping up in the industry. I wish there was a bit more alignment across companies because I probably count 4 of them at the moment and I don't know which one to embrace, but I can see a future in a few months where most of our agent usage will be driven by agents, not by humans.

>> Yeah, and the harness is a big topic. Harness selection could be a big deal. Yeah, especially in the enterprise. Cost per tokens. I mean, cost for the models. I mean, I just saw a study this morning from the CEO of Databricks, Ali Ghodsi, and he— they did their own internal benchmark on all their source code and they found that the models are pretty much the same in price. It was harness selection that drove it. And that was interesting. I didn't see that coming. That points to some of the nuances around backend. And so that works, then the frontend platform enablement happens. And so that's a good thing. What about harnesses? You guys take care of that for users, or is that on the enterprise side?

>> We bring our own harness. You know, we've been putting more than 2 years of research on that, and we, we own the harness for 2 reasons. First of all, we build it for non-technical users, so our harness is more opinionated, more autonomous, and really does its best not to put you in front of technical decisions, because those are the decisions that become blockers, and that's when you stop using the product.

>> Yeah, yeah. Which MCP server should I use?

>> Exactly. Imagine someone in marketing in front of the question, it would just close the tab and disappear. But the other reason why it's also very important to own it is because when you want to build an harness that is capable of interacting with the entire platform, then the, you know, the outreach is much bigger than a pure coding agent. So I don't think of our agent anymore as a coding agent. It's literally like an agent that is capable of using the entire platform, you know, correctly.

>> Yeah, it's a building agent. You build, you build products.

>> Yeah, it's exactly the entire software development life cycle. Cycle, not just creating PRs for a certain repository.

>> All right, so what's your goals now? You got the growth strategy, got a billion dollars you're eyeing for the revenue. What are some of the goals that you're working on? What's your focus? What are you optimizing for?

>> I'm focusing on agents, and what I mean by that is not the Rapid Agent. I want everyone to be capable of creating their own agents. If you think about it, it's such a massive unlock to be able to create an agent. It's literally what drove our success, and I want to bring that capability for everyone.

>> Everybody wants an agent. Who doesn't want an agent to work for them?

>> I build them for my own work. I'm already overwhelmed with information. If I didn't have them, it would be even worse.

>> Check all my email for me, tell me what I gotta do.

>> Delegate tasks at the right time. But agents are really hard to build correctly, especially with all the governance requirements we talk about and reliability. And there are very few companies in the world that have experience to do it in depth, and of course we are one of them. Our take, of course, is always very aligned with our company mission. This is— this should be a technology that is in the hands of everyone. So the next big launch we're going to be talking about, you know, maybe in a few months, invite me back, you know, we can do it.

>> I'd love to.

>> Yes. Yeah, same revolution we brought to app building is going to come very soon to agent building.

>> Are you going to have that preview on the consumer side first, or the individual side, or enterprise at at the same time? Because individuals might not have a governance issue, or small business might not have a governance issue. And this is where I think it shines, because if I have everything in Amazon, I got one governance policy already set. Is that going to be the first place, or are you going to do it all together?

>> I've been thinking about it exactly in that way. It's always easier to ship things in front of consumers and then learn the lessons there and then bring it to the enterprise. The difference between when we launched a couple of years ago is the fact that Enterprises now have visibility on what we're doing and they're so eager to get new features. So I'll do my best to make those launches almost overlap.

>> Yeah. I mean, you have to do the enterprise. One is great market into what you're doing, democratizing the creation. But also there's a lot of shadow AI going on.

>> Yeah.

>> You know, like it's like shadow IT with cloud. We saw that before. There's tons of it. The finance teams vibe coding and coding up stuff. Stuff can break. You got to rein that in.

>> Yeah, yeah, yeah.

>> Okay, final question. RACE Summit, obviously you guys were featured. I saw on the schedule AI infrastructure, a wide-ranging AI conversation up and down the stack. You guys obviously at the front lines of that on the frontier side. What's your take of the show this year? What's your opinion? What's your summary? If you were a reporter, how would you report this show?

>> First time for me at the Race Summit. I really enjoy the fact that it's full stack, as you said, you know, you get lost across the rooms and you see all the way from the GPU buildouts to the app layer like us. So great to meet founders across the entire stack. I'm European. I'm very proud to see like a flagship AI event happening in Europe. Paris is an amazing city. It was also fun for me to be like on a flight from SFO to here and meet Alfred DeFalksen on the flight, one of my friends in San Francisco, you know, like all founders in my space. It's great to meet you.

>> The whole tribe was coming from the Bay Area. You were at business class, all the business class seats were taken.

>> Yeah, but I think all of us are here especially to meet companies that are not exposed to what's happening in the Valley, but they're very eager to understand and embrace this level of innovation as soon as possible.

>> Yeah, and the show had a great success, the event on the opening night was pretty impressive. Yeah, Palace of Versailles was great. Fireworks, great stuff. Well, congratulations. We'll definitely want to follow up on your news. Looking forward to it. Thanks for coming on and wrapping up our show.

>> I'll be back.

>> All right.

>> It's an honor.

>> All right. We're wrapping up. This is a wrap for Ray's Summit Day 2 closing. I'm John for your host, Cube. Thanks to the team, guys. Great job out there. You got Gemma Allen, Alicia Volante, all doing a great job. Liam, all doing great work. See you next time.

>> Good.