All right. What's up everybody? We're back. Welcome back from our commercial break. You see Gazi has transformed into a hot over here.
Same haircut.
All right.
I don't know. I'm working on it. It's getting a little thin up here. What's up everyone? So we were just chatting with Gazi. Now we got AMD up here. He's making a little post on X. He's getting on his phone here. Hyp up a little bit. And did you just get off stage? Are you going on stage?
I did. I did. I did. It was a very cool format. It's sort of like a talk show host format. He had this desk in front of him and his iPad and computer and he was going through the stack that runs Saster. So for people who don't know, the entire SaaS experience, not only the website, the back end, the database, but also the marketing, the sales, the sponsors, all of that is run on Replit, either through Replit apps or Replit agents or agents on top of Replit. And it's fascinating. So he was running this conference. He's been running it since 2015.
He used to need like 20 people, 20 plus people, a bunch of contractors, a bunch of agencies and things like that to build a website and all of that. And now it's just two people, him and Amelia.
And a bunch of a couple of AI agents. And they're doing this interesting thing where they want to hire the first person in a long time and their idea is that person reports to the agent.
I don't know how I feel about that. We'll see how it plays out. But he's like, you know, I ask our marketing agent, it's called 10x 10, it's 10x 10k. And he's like, I ask it every day, 'What should I do today?' and he's like, 'It's giving me, it's a better manager of me, the CEO of the firm, than I manage myself.'
And so then he had the idea of like, why don't we hire someone to help it, the things that it struggles with. And so, it's really, Jason and Sasher, they're living in the future. This is, I think, the first semi-autonomous event. A lot of the decisions that were made here were made by agents running on Replit.
Yeah. A lot of the people are here because they were sold and marketed to by an agent. A lot of the sponsors, the entire sales cycle with those sponsors happened on an agent I think called QB. And so we went through this entire stack and he got my views on it and at the end I was like, 'Dude, you're living more in the future than I am.'
Yeah.
Can you talk a little bit about his background, too? I feel like Jason went all in on Replit like maybe 6 months ago. And it's important to note that he doesn't have a software engineering background, Amelia either. I think there's just like entrepreneurs that know what they want and know how to talk to AI.
He's an entrepreneur. He's a VC as well. He's a very successful VC and he's a community builder as well, but I don't think he's written a line of code in his life. I might be wrong about exactly that, but I know for a fact that he's not like a coder, professional software engineer. And he started using Replit I think around Agent 2. And man, things were so different back then. It was like almost a year ago when he started, I want to say like April, March 2025. And the agent was kind of primitive. I mean relative to where it is right now with autonomy, with parallel agents, but also the platform wasn't as mature. I mean we matured the platform night and day. The platform is a lot more scalable, a lot more secure, a lot more reliable, like deployments fail, go down way less. The uptime on Replit is amazing. Like the uptime of Replit apps hosted on Replit is better than Replit.com.
We should maybe host Replit.com on Replit, which is something we want to do at some point. But so I think when he first came on he struggled a little bit, like a lot of people I think in 2025 started using agents. It required a lot more grit and a lot more technical knowledge and a lot more just perseverance to get things and he has that perfect personality. I think a lot of people actually from that era, there's a lot of people that kind of tried these products and you talk to them today it's like, 'Ah, they don't work, you know, put in a random prompt and very vague prompt and it didn't give me what I want and I just left.' But he persevered and he ran into some issues and he overcame them. And I think by October 2025 around Agent 3, the kind of things Jason was doing with Replit was far more advanced than any software engineer was doing with agents.
The thing that surprised me about the hype that started around December 2025, especially with software engineers, is that I was like confused because our users are already doing exactly all of that. It's like software engineer was like, 'Oh, you don't have to type code anymore.' It's like, 'Well, our users have been doing it for at least three, four months now.' And so, I think Jason through his persistence was able to become a vibe coder par excellence.
How does it feel to be a little bit at the head of the market in that regard? Because it seems like that happens a lot, right? Like I see people getting hyped up about announcements that come out and I'm like, 'Hey, we've been doing that in Replit for like six months or whatever.' Like how does it feel to kind of be making decisions that are leading the market and how do you kind of go into those decision-making processes as you guide a company like this that's growing so fast?
Yeah, I mean look in some sense we are a much bigger company now. We have this large valuation, large revenue and run rate. We're closing on 300 people. But in many ways relative to our peers we're way less capitalized. We're way less hyped. We're not on TV and on the New York Times every other day. And so we're punching way above our weight when it comes to innovation. And you know, I can tell you for a fact, every time we release a new agent, I hear from my friends at these big labs, they all huddle around the computer and look at what we're doing. And when we release especially Agent 1, we've gotten reached out from AI researchers that told us, 'I wasn't aware that LLMs are capable of this.' So it feels really good to be able to kind of lead the way and show people what's possible. It often times feels a little frustrating because you're sort of at the edge of what's possible. Technology breaks in all sorts of interesting ways, but also when people finally catch up and it's like, you know, we're already doing this. Like, you're just not paying attention.
So, but it's great and I think as a company, as a culture, we pride ourselves on being first on so many things. Replit was kind of the first online IDE, the first IDE that bundled together deployments and database and all of that. There's so many firsts even like simple things like every IDE had a save button. Replit was the first IDE to not have to do autosave, right? So like there's a lot of things where you know one of our values at Replit is we're famous for the seek pain value but there's another one less famous is called think radical and the idea is at any given point in time there are conventional wisdom that is outdated that you can actually rethink. Now, often times conventional wisdom is right, I think, 99% of the time. But if you're able to find that 1% of the time where something is no longer serving a purpose and you were able to rethink it, that's how you can build great businesses that grow fast.
It took me months to get used to the whole autosave thing in Replit. I have to continue, I keep doing like Command S, Command S and give me something. I was like, oh that's right, Replit automatically saves everything for me. I don't have to do that anymore. Yes. As you're talking to power users like Jason, how does that change your view of the Replit experience? Like he's living in the future as you mentioned, how is that impacting how you're thinking about the roadmap for the product?
You know, the main thing when I was on stage kind of listening to him talk about all these things that he's building is like we got to help more people get to his level. And I think the things you guys are doing and the things that the rest of marketing devrel folks do with educating people is very important but I think there's more to do in the product to guide people. We've had a couple of false starts to help people build agents. We have this automations product that I don't feel very proud about but we tried a few times. I think the next big thing we need to build is really in an on-rails way. We need to do what we did to apps. We need to do to agents what we did to apps. Before Replit making apps was incredibly difficult. Now making agents is incredibly difficult.
So we need to figure out how to make making agents fun, reliable, testable, deployable and scalable and easy to build.
You have your context all in one place and this artifact is an agent rather than an application.
That's right. That's right. And I think that's the future of software.
Well, and we're getting really close. I mean, we've got parallel agents kind of running and that's like a major, I love being tied up with the main agent and I think of something else, I'm like, start a new task, send it off in the back. And then we have the little feedback button in the bottom which sends the stuff into the agent that you can kind of feed. So, we're not that far from like self-improving tools and kind of the agent taking the next step.
The main thing is to platformize this if that's a word. But like all the stuff we did for Replit Agent, I think we're one of the best companies in the world at building agents. And so can we take that knowledge and make it into a platform. So can we provide the same platform we use internally to build Replit Agent for people to build a marketing agent, a sales agent because all these agents have the same architecture ultimately and so creating a platform to support different types of agents and then them inheriting parallel agents, compaction, context, you know, all the stuff we do, skills. So yeah, we just got to make it easy and we're embarking, we're now doing like this big rewrite to make what we built into a platform that others can leverage.
Interesting. There's a big step change that's happening for companies. They're going to have to rethink their offering and their services. I guess I'm wondering what advice would you have for those companies as they're adapting to the new AI landscape like agents doing more work but also changing their value proposition. Like what opportunities, what pitfalls do you see for founders?
I think a founder today needs to be an athlete. Needs to be. And when I say an athlete, like if you think about push-ups, definitely push-ups.
Inside, sorry. Buildathon joke. Somebody built an app that scanned your body and to test it. We put the app up and then I did push-ups on the stream and it mapped, you know, and did the...
He got up to five. Wow.
Which was great. Great. I mean, I'm sure a lot of people will take this literally go build fitness apps. That's not what I'm saying.
Um, when you think about athletes like I don't know Michael Phelps or chess, think about Magnus Carlsen. Magnus Carlsen needs to be on his A-game every day, like every day. You can't, you slip up and you lose against Hans Niemann and you accuse him of cheating. All this chess drama, but you have to be ready all the time. And I think entrepreneurs and people working at startups need to, the main thing they understand is that you can't skip a beat.
You need to be on top of what's happening in the AI world. You need to be in touch with the research, with the latest capabilities, with what's possible. You need to be constantly building and you need to constantly reinvent your product. And one thing that is important I talked about on stage is the sunk cost fallacy. You know this idea of like oh I already put so much effort into this you might as well invest a little bit more into it which is why people don't pivot at the right time or they don't change the product in the right way. And it's important to totally cure yourself of sunk cost fallacy. I think it is one of the main killers of companies today where they're like, 'Oh, we've invested so much in this platform.' Well, this platform is outdated now. Like, you got to renew it. And so, we try really hard at Replit to be very okay with starting fresh.
Yeah. And it's not even about just being the athlete. They obviously have to be ready too, but creators or vibe coders or people who want to kind of innovate in the space also have to be marketers. We talk about this all the time. They have to be media specialists. They have to be monitoring the industry. Like I work in the space and I feel like all I do is try to keep up with this industry and I can't keep up, you know? So like what else do people have to think about if they want to be competitive in the space? Obviously not building at Replit scale, but most of the people who are watching here are trying to build their freelance business or build an agency or get a product off the ground and like, you know, what are you seeing from, you know, to be on the cutting edge to get to the next level as a creator using a tool like Replit? What do you see as that kind of marquee skill set or an advantage you might tell somebody to focus on?
You know, I phrased it on this podcast with Jack Neill in an interesting way that went viral. A lot of people are upset about it, but I'll say it again. I think there are things that used to be vices that are now virtues. One of them is being lazy. And what I mean by that is I think lazy people are good automators because they're like allergic, I am allergic to repetitive tasks. Like here's a Replit app. Every time anybody asks me to do anything even remotely repetitive, I'm like, 'Replit agent, can you solve this for me?'
We have a chat scraper right now that's looking through all the comments and that came about because I asked Rhymar once to get the comments from YouTube after the show. This is a Replit built app. So, he built out a Chrome extension that scrapes automatically. Now, we have all of our comments here.
What's crazy is now we have a whole live and we can't show this. I wish we could, but this is the comments. We're building like a whole tool to start interacting with folks. The agent will start scrubbing for questions and building a queue. So like and this is tied into the Ream infrastructure too. So like the tooling that we're building. One of the things I'm so excited about is the stuff that we're actually building here in the community using Replit. And I think we're slowly starting to kind of reveal some of that to the community and people are going to start to see like oh this is really cool. Obviously this buildathon platform that we just did. We just did the 24-hour buildathon and we did that on Replit like we built that stack on Replit and it was kind of cool to see because there was a point where the app crashed because like the traffic for the buildathon was like so big. We had to have engineering come in and crank up the servers and then we had to do some real-time optimizations but within 15 minutes of this spike that we got everything was leveled and balanced. We published and everything was great. Shout out to Tanner in the audience if you're watching. But I don't think any other company would have trusted their own stack to do something at that scale. You know talk about the confidence that we have as a company letting people build that type of application on a tool like ours and hit that type of traffic and just confidently know like hey we're doing this. We're not going to go to some hackathon specific post like most of the other competitors are doing.
Mhm. Yeah. I mean, you got to be, I guess, honest or you got to be not hypocritical and you have to, they call it dogfooding. I don't like that term, right? I think it is steak. It's not dog food.
Drinking our own champagne. I've heard...
Drinking champagne. It's only champagne if it comes from France, right? From the little pocket sign. I don't know.
