SaaStr Day 3 - What's Next for Vibe Code

Are you ready? — drop. — Heat up here. Heat. — Hey. Oh, I thought duplicate. — Hey, what's up? — What's up everybody? Welcome back to another Repid live. We're here day three at Saster and it's been an exciting few days. Uh, you know, exploring the AI landscape, the entrepreneurial landscape, kind of seeing all the different companies come up, speakers on stage. We've had, um, let's see, we had Gazi, Amjad, Miy, we had a couple people from the crowd, firsttime builder, a little bit of experience builder. We had Cody, we had Jason Lmin. Um, who else did we build from Green Security, Julia and Lindsay on the show yesterday. That was really good. — Well, and the amount of alpha that's being dropped on this show that like not that many people are kind of um seeing compared to like our normal live streams that's going to get clipped up out of this. So, go check out the especially the uh I'm Jot Michael and Gazi interviews from day one. Um, yesterday was more like thriving in the day talking a little bit about like how this stuff's being used in real time and today I think we're going to, you know, just wrap this up a little pin on top of things. So, how you feeling, Manny and uh, Francisco in the game? — Great. I'm doing great. I just wanted to huge shout out to Nani, KG, Dario's chat. Dario, I need your time zone. We're going to get on a call soon. Uh, summer camp, Raz, I've seen a couple different people jumping in and again, thank you so much for joining. I mean truly these are, you know, built out for you and I see that someone needs to catch up on four days worth of content. So, we'll do our very best today. — Yeah. Do a quick recap. Well, we're definitely moving faster at faster. That's been the theme. Oh, rewind. — Faster at faster. Um, no, I mean I think No, it's just been great. I mean, I think um we've been talking on the streams around VIP coding going pro, vibe coding being used more in the enterprise on the business side and so here like everybody's like all in on that and so it's been great to see people are actually building. I feel like the commentary has been like 12 months ago people were curious about what AI can do but there's still a lot of open questions about how to bring it into work, the business. And this time around it's a lot of handson building like the vibe coding sessions that Cody has been doing. Horatio has been doing — Jacob Eisner as well. He's going to be joining us. — That whole tent that's like a whole

setup there. Like they're killing it. — Bring up some if I can pull that up. — And yeah, for for those of you that are not here that are joining us. Yeah, it's like a big campus and there's a bunch of open air um areas as well. We've got a vibe coding workshop outside where we have replet folks building here. Um, we'll bring it up here for you. And the weather has been awesome and these tents have been packed. We came in yesterday pretty early. Didn't we come in early yesterday and this place was packed. We had Horatio uh showing some examples of what to build, taking some questions from the audience. We had we've had Cody here do a lot of great builds. And so I think yeah the the key takeaway that I'm getting from this year Sasser it's time to build now you have the support coming from the top down executives and business leaders are saying we got to do stuff with AI they're empowering their team they have a mandate we've continue to hear that like every company now has a mandate we got to get with this whole AI craze and then also now you have the tooling that's coming along with that that's making that possible so folks can go into replet and start building stuff and feel secure about doing that and feel safe about building there. And so you just have a bunch of things that are coming together where it's like a lot more building that's happening and I'm really excited to see that. — Yeah. No, I am too. I mean, it's been absolutely incredible. I mean, I'm meeting a lot of people out here who are building for the very first time and kind of just really getting into the, you know, weeds of it. And it's been great. I mean, I think that's why these sessions are, you know, filling out. People are really trying to adapt to this new, you know, world of AI. And I never get tired of seeing people's impression when they use Replet for the first time. Like I call it the Replet glow. When you like get onto Replet and you build something and it's functional, it's cool, it's neat. It's usually in a one shot or a couple of prompts and people just are buzzing with that feeling with that excitement. And if you want to see a glimpse of that, check out yesterday's show where we had Julia and Lindsay join us from a company called Green Security that is early in their AI adoption journey and they just started using Replet and they're just raving about it like oh like look at all the cool things that I can build allowing me to go faster and the company's supporting us and they're really excited about building with Replet. So I don't know I don't get tired of seeing that. I love that replet glow sensation that I see off of new builders. I think it's one of the things too I also love is you get that initial kind of replet glow, but then the follow-up a week later. Here are the six apps I created and here's what the next iteration is. I love it. — Well, it's I think it's the replet glow the first time you use Replet, but then like fast forward two weeks, I call it the replet like bags underneath your eyes because you've been up all night working on your replet project like one more prompt, one more feature. Like you know about that, right? — I don't have no idea what you're talking about. Like when Rhymar and Marcelo got free credits, like I just started getting messages from them later and later in the night just with random updates. It was awesome. — You see uh you see Rhyar status always green on Slack cuz he's like always on there like just jamming away on Replic. — Yeah. Yeah. Y'all are pretending like I wasn't a replicad before, — but now we let you into the Candy store, dude. Crazy. — No. That's right. That's right. It's all over now. — Um yeah, it's been really cool. Well, we're uh I'm we're showing some footage of the the conference floor here and just kind of running through the experience a little bit. Um these are the ladies that we had on the stream yesterday. — Um — yeah, Manny evangelizing. — Is that audio coming through? — Yeah, Manny doing his recruiter mode here, trying to grab people from the floor. So, again, pulling people in and just chatting about what what's going on. And um — yeah, I don't know. Um, this is one of those, again, I go back to like maybe you're not here at Saster, but there's opportunities for you to like get connected with people in your community to do this. Like the again, it's infectious to just do this with other people, even if it's one or two other people. I remember some of like my early meetup days, some of the best meetups were just the three or four or five people kind of getting together because you can really nerd out on this stuff. you can really get uh hands-on with each other and there's nothing that beats kind of being in the same room and learning with other folks and getting again hands-on with uh these types of things. So anyway, yeah. Um and I don't think you have to do much like I feel like I keep saying that Replo Replet Replo — what's Replo? That's the next new brand. Yeah. Talk to Philip our branding guy. Like I got a new name for us. Replo. Um I I keep saying that Replet demos. So well, like it's one of those things where you can have like five minutes of commentary like this is vibe coding. Here's what it does. It allows you to build software even if you're not a technical person and then just break out your laptop and put in a prompt into Replet and just wait five minutes and let Replet cook and it's like oh everybody sees the magic of that. So you don't need a lot to go out and hang out with folks and show them what you're building and introduce them to Replet. And when you introduce them to Replet, they're just going to thank you forever. And if you do, again, I

keep saying this, make sure you're using your referral code in Reply. I keep seeing a lot of people going out there sharing Reply. Here's what you go into your settings and you find that URL that says that if somebody activates on that URL that you get like $10 off or something, whatever. It's a great way, you know, get a little something for your for your token um balance, but also it's a great way to track how many folks that you're introducing to Replet that are actually converting into real Replet users. — Yeah, absolutely. And as you, you know, use a referral, always let us know. We're always looking to see, you know, what folks are building out there, who are they sharing it with, and there are a lot of events and it's just only ramping up. So, we have an event next week in London. Uh we had, uh, an event Toronto uh, last week. We had Lima, uh, Ghana. There's just tons of cities that are really getting activated here. Um, always check out our community homepage for more information on upcoming events and just general updates. Um, and yeah, and then, yeah, let us know what you're building out there. We always love to see it. A lot of the folks that are part of the ambassador community today were just people that got up and get started. — Here's another key theme that I've been seeing is around folks building more agents on Replet. Get some context here. Jason Lmin, the organizer of Saster, he's a big-time Replet user. And what he's saying is that this whole event has been put on by him and one other person using like 30 agents all built in Replet for like the marketing part of things for the scheduling part of putting together this conference and a whole bunch of other things and like the cold emails that are going out. He joined us on the live stream yesterday. That was really good. Go check that out. So he's all in on agents. And then we also had Amjad on the show on Tuesday and he's just talking about how like more and more people are building agents on Replet. So um become the nerd. Yeah, you're definitely on track here. You know, I keep thinking to Amjad's comments around like us making it easier for you to build standalone agents in Replet. He's seeing a glimpse of the future. He's seeing what Jason is doing with agents and we have some way of doing that in Replet, but his view was like we have to make that a little bit easier and more straightforward for you to just go straight to building an agent. So maybe you don't want to build a website or internal tool or an automation. You want a standalone agent that might like help you do some of this stuff. — Yeah. So, let's anchor back maybe to the uh what's next for Vibe Code topic here because maybe we can — uh future cast a little bit based on our inside knowledge based on being on the ground floor. Maybe we project forward like uh you know what's going to be — this industry and maybe you all are watching this and you have some thoughts like drop some thoughts — where does this go next? Uh what comes of all of this, right? uh — Sasher 2027 — fully autonomous. Are you automating your workflows? Are you building different agents to do things for you? Uh and then where do you think these capabilities are going? Uh that's kind of what we want to know throughout the stream today and going to be the running topic is like what do you see next? Right? We've kind of talked a little bit about here's what's going on now. We had Amjad uh Gazi and Michael on day one kind of talking about this is the current state. This is how we see it. Here's how enterprises are adopting it. Yesterday we talked more about, hey, how do you thrive in this space? Here's some new builders. Here's kind of what's happening right now. And I think today it's like, all right, we see where we're at. What comes next, right? What do people need to start thinking about as they see this now? Right? Like people are looking and they're starting to say, what is this? I've heard about coding for a year now. I see people doing things. I hear people talking about it. I say I see haters saying one thing. I see the evangelist saying another. That means the truth is somewhere in the middle. So, where do you all see this as like the next step? And I think maybe that can guide uh part of the conversation that we're planning to have uh for the next handful of hours here. — And then we'll be able to grade ourselves one year from today when we're here at SAS 2027 like which takes really played out. Yeah. Who was right — poly market the the Yeah, we're going to have to build a community polyarket here for like these types of predictions and things and uh maybe like buildathon winners. People can wager on buildathon winners inside a poly market, but you got to spend your replica cash. I was thinking, okay, total aside, — side note, — side quest, — my YouTube algorithm uh feeds me, you know, in the shorts, this uh this guy that just like he and I've been watching him forever. It's like he gamles a certain amount based on the number of followers he has. He was like one of the first people like there's lots of people doing this now but this guy like he's almost at 2 million followers and he's like — and anyway there's like this people just follow for what I just keep thinking like what if there was like a replicash casino like where people could take their buildathon earnings and just like you know spin the machine you get on the board and then next thing you know like — what do you mean but he gamles by the more like users — bro so he has a certain number of subscribers and based on the number of subscribers he makes bets up until that amount. So he had likeund one and a half million subscribers. So he had like uh $150,000 that he was gambling on the

blackjack table. So based on how the subscriber count grows, he kind of increases the wager. — Uh — and is it the delta or like the absolute level? So like is he betting like 200k every number? So I think every subscriber is like uh like a dollar or something like that. Uh maybe not a dollar because that would have been a million and a half, right? So there's some kind of translation where like he's betting, you know, $150,000 — like every time, right? — So he gets 200. — So when he hits 2 million, he'll do 200,000 as a bet. Yeah. So anyway, there's like this increasing and we talk a little bit about this audience engagement. So like I just kind of watch sometimes I'm like cuz sometimes he gets cleaned out — and then sometimes he Yeah. Anyways, side quest over. — We'll gamble and if we win then we'll like give it to somebody in the audience or something. Anai talked a little bit about some of these — I don't want to say degenerate tendencies — but some of these things that are looked at as like not necessarily positive traits in some in some circles right yeah but that those kind of become advantages in this world where it takes like this hyperfocus and this ability to jump between projects and like kind of this uh really diverse skill set you need to harness to kind of to be a part of this and so again we're kind of triggering some of these dopamine things and I think we're really early in the psychological evolutionary process of how these all these digital things affect us as like a species. — And I think what we're going to find out about the early social media days is that it was just a little predatory generally. And I think we're just on a better path to figuring out like how do you know again trigger some of these reward mechanisms but then result with productive outputs rather than just like brain rot. And I think that's what we're starting to see that a little bit, right? And I don't know if this is the perfect version of it, but for some reason those two things and I don't know. I just — So is that your forecast? Is that one of your bets? Uh, I do bet that lots more people are going to replace doom scrolling with like upskilling and that entertainment is going to become uh I'm sorry, education very much more educational. — Edutainment — educ education is going to become very entertaining. What am I trying to say here? — I think you're trying to say edutainment, right? That term where like you combine learning and fun in a cool way. — Well, and the gamification, right? Like the rewards, the incentives. Right now the rewards and the incentives around the internet are basically like hey how big an audience can you get and then can you uh pervert that audience for like advertising revenue whereas I think what we're seeing now is like hey can you generate attention but then can you use that for something productive in the real world and that's what I think we're seeing these new networks we talk about these action networks versus the attention networks and so that's what I think that would be my prediction is that we're going to continue to see people build new things with these tools that allow them to activate their small group their niche there something other that that's just kind of been floating ethereal on the internet and kind of harness that towards some kind of goal. That doesn't always have to be money. — That could be passing some local initiative or getting uh some kind of nonprofit funding or whatever, you know. Um but I think we will see a lot of that and then localized tools. So now you're not beholden to some software company for hundreds of dollars a month for this like one little sliver of a feature you need. It's like we can do it, you know. — Yeah. — I do see that. That's that's that would be part of my prediction for sure. — And KG is a great one as well and I just keep seeing this more and it's like point andclick UI is going away the way the dinosaurs voice control just in time generated UI. I mean I see this all the time in the office, right? — Where it's it's an interesting corner over where we are because everyone's talking to themselves and of course we're all just using whisper flow. We're talking to our agents, things like that. I think there's also a very much a shift coming in how we interact with our machines. — I think it's a good take um KG. I think the UI in terms of how you interact with applications is changing you as an individual user. But we also had Jason on the show yesterday and he's talking about how like the way that our LLMs and our agents work with products and services. That's going to change. He's like the number one thing that companies have to be thinking about right now is how to make their offering agent friendly. So for example, he's not necessarily going into Salesforce anymore, but he has his agent going to Salesforce and he's getting the he's getting more use of Salesforce than ever before. It's just not him actually going into the UI clicking around and finding things. He's just directing his agent to tap that data in Salesforce and whatever services they have on there and then bring that back into his agent hosted on Replet. So this idea of a headless product where like the UI necessarily isn't like the area where people are going to get the most value. It's going to be around making it easy for agents to interact with that platform. — Yeah, — that was a really hot take. That was really good. — Yeah. Uh, I would push back a little. — Oh, tell me. Let's do it. — Um, I think for sure we will have point-and-click real-time UI generated. But I also think uh the cognitive load of like trying to figure out when I open a new app that does the same thing, — do I have to relearn the UI? — Right? Like so if you look at websites and you look at like a lot of standardized pro like there's a lot of things like oh the logo is probably in the top left you know like the nav's probably in the top right you know like

oh it's an app okay so maybe there's going to be this little sidebar on the left you know it's like some of these things are just cues that I don't know that we want to get rid of or let me put it this way you get rid of them at your own peril as you create new software and people want to explore that and then you make it more difficult for the cognitive load for them to figure out where they might want to go. So for certain things, I think that's probably right. Right? If I want just a summary of something and I want to ask the AI and it's just in time and it's just like it's creating a feed for me or it's creating something simple. But if it's something that I kind of want to access regularly and every day and it's like I kind of want that to be consistent and then probably we want other people to have consistency across the board because if you and I are in the same software then and I need to explain to you where to go somewhere then I need to be able to direct you to that specific place and it shouldn't be different for me than it is for you. And so I think there's probably like truth on both sides of that where yes, a lot of things will be generated on the fly, but I think just from the mental load and from the psychological aspect like people are just going to want familiarity in a lot of places. And so I still see trends and patterns in UI kind of — even though the AI may be generating stuff, there will be some basic framework for navigating and understanding these things. Otherwise, it'll just be very difficult. — Yeah, I get that. And I guess I would say um I think it's going to change radically what that UI looks like cuz to KG's point like now like I'm yapping a lot more to my computer more than I'm actually kind of obnoxious at the office I think cuz I'm like oh like no change that add that to the and I'm like talking to my computer all the time. — So I think like that has to get incorporated into how we go into like a Salesforce or like a data bricks and viewing the UI. I think there's going to be more of a change there in terms of when I do go to the UI. want more of this way that I'm working today. So, I think you're right. There's probably going to be like some trends that or some patterns that surface and everybody adopts. I just think it's going to be different from what we see today. — Yeah. And I think what you're saying, too, is like instead of just navigating through the traditional UI, you should also expect to navigate through natural language, right? Just like you can talk to the AI, kind of ask it things and it has some kind of intelligence or ability to like interpret that. But that doesn't necessarily in my mind mean shifting the UI. It could maybe, right? Like maybe you can optimize it. Maybe you can and you could do this with browser extensions now. You could just like, hey, I want to change the color of this background and like I want to customize this for me. So we might see more customization or hey, I want the sidebar to have this element in it. But everyone has a sidebar or a thing. And again, it depends like are you building a dashboard for yourself? Is this a thing that you just expect to use? and then it's fully like it makes perfect sense for it to be customized fully to you — or is this thing a million people going to use it and you kind of have to standardize it because you have to deal with a lot of different — uh you know mental models that need to be unified and so I think again the use case will define whether or not it's like super custom or a little bit more standardized across the board. — Very neat. Um hot take not hot take but prediction number two that I have. Um, I think that we're going to see the rise of the oneperson team. This was something that was coined by the CEO of Coinbase, no pun intended, um recently around just how Coinbase is using AI more and how AI is empowering teams and individuals and this idea that you have one person that's doing everything involved that you know getting an idea like started and then like actually shipping it because now you have AI that can help you do the research build the prototype that can help you assess how people are using it and then like maybe even ship it as well. Like I don't know if we're getting to that point where people like new vibe coders can ship production grade code. I mean I think in some instances but I think still a small fraction but more and more I think you're going to see this term the one person team that does all of these things that usually required five different hats to ship new products. — Push I'm going to push back. Okay. So, I definitely think uh we will see lots of oneperson teams doing lots like shooting way above their weight. — Um, one of my favorite phrases is if you want to go fast, go alone. far, go with a team. — Right. — And so, I do think there's going to be a lot of people jumping out of the gate and being able to like this freelancer model, this independent contractor model, you're going to be able to do so much more, right? As an independent contractor, as a freelancer, you will be able to compete with agencies. is basically have a team with you of these people doing you know some of that stuff we talk about what the AI use cases are good for and what the agents are good for. However, I still think that having other people again we talk about this human side. We talk about the connection side and I'll make this other point. I keep going back to lots of apps that we saw in all of the buildathons which were check on your friends. What's the status? How do you keep up with people? How do you connect with like, you know, these people that are in your network that maybe you're losing touch with? I saw so many apps around that. And so part of me thinks that there's this trend moving forward that we're seeing emerge and people trying to build these apps. What that's going to lead to is the real connection. So I think yes, you're right. Individually as creators, we are going to find like more capability in our creative skill set.

What I think — we shouldn't forget is that working with other people in that same direction makes you better. And so what I think we might see, this would be my prediction in a vibe code. Not to counter yours. I don't I think yours is probably right. Um but I think what we — Oh, he's hedging. — No, no. I'm adding. It's like a Yes. And — I got you. — What I think we will see is a new type of like corporate structure form, right? Uh whereas — corporations right now and businesses right now are very hierarchal and they're very top down in the way they're structured. What I think we will see is networks that are loosely organized groups of these independent freelancers that do cool things and they come together to kind of like be assassins in one specific point, right? So, it's like you need this creative project done, you're going to pull boom boom and pull this team together. They're going to go do this cool thing. Everyone's going to get paid and then they're going to disband, right? And it won't necessarily be like the umbrella that they operate under will be some kind of community, but there may be like shared buying power for things like benefits or uh leveraging negotiating contracts because we're all going to buy tools together or spend AI credits in a certain place or all we're all going to focus our attention in certain ways. And that goes back to this concept of like these action networks that necessarily we don't need to be one single entity. We can all be individual entities, but we can work together in some way to push an idea forward. And that's where I would, you know, venture to say that like yes, individuals will be stronger, but you'll always a group of those individuals working together will just, you know, that team will just be amplified. And again, that's where I think like we're showing the way of how what does a team look like with AI forward. That's why I keep going back to like regardless of what happens with these tools, with, you know, AI on a longer term scale, what we're seeing right now is an AI enabled team here at Replet where AI is a first person citizen. Yep. we're using these tools daily to kind of amplify our work and so yes me as an individual is maybe doing the work of two or three people on this team because of these tools right and I think that's a new paradigm where it's like maybe we don't have less you know we still got to have teams but those teams will be much more amplified to your point — yeah and the way I would frame I mean you're a great example of this where like you can go really far on a project and you're still bringing me and Francisco in like you're building the buildathon site right like hey do you guys what do you guys feel about this you like this give me some commentary but you're driving most of that development and you're still bringing us in to give you commentary. I think that's going to be like the more of the model moving forward. Like one person can just drive hard on a particular outcome, just keep pushing on that, but then you also want to bring folks in to give you some advice, give you some guidance, make sure you're like you're um like getting people keeping people up to speed on things. Um, but I just feel like one person with AI is just gonna be able to just march on something and just like get stuff done — to push again like but what you're saying is like let's say so Rhymar's focused on building the buildathon site but then you're focused on let's market the buildathon site and then Francisco you're focused on like okay on a high level how does this fit into the strategy and then we have to tie in these other teams and now it has to align with all these other initiatives and so it's like — yeah you're right — yes you know each individual is amplified in their way whereas maybe this might have had to be a 10erson team yes — to run this buildathon and we ran it with four or five people with the help of some support of other folks right as needed. And that's where again I think like — individuals will be able to just pull a little bit more weight in the direction they're pulling, but having multiple people pulling in that same direction is always going to be better. — Yes, — we see that in our own team, right? I mean, the stuff that you're doing on your own and just building out and constantly developing based off, you know, the community feedback has been unbelievable. Manny just like the amount like all the content you see online a lot of it is being produced by Manny right like being pulling together things you know creating things obviously Rhyar is creating like assets but again — I think individually we're doing like multiple different you know jobs at once but we're all headed towards you know the same community — uh goaling of supporting all of you — yeah but versus like two years ago like think about like when we do a buildathon like we do streams the assets the site itself the support the grade the scoring I that would have been like a 10 15 person team to do that and you know it's like us three kind of like hammering that out. So that's like my next kind of um prediction here is that you're just going to see leaner teams. You're going to see flatter organizations. You mentioned that we're seeing that at Replet. We're flattening the organization. You're seeing that at other companies just leaner um flatter organizations that just can move fast on things. — Yeah. I [clears throat] agree. And I and this doesn't mean that teams that exist are just going to shed all their teams, — right? We talked about this with uh Michael and Amjad also about some of the downsizes happening and how some people are just kind of like uh AIashing their layoffs as they restructure. And what I think is happening more than anything is not like, hey, these jobs are being removed because they're no longer necessary. it is we need to reorganize the team in order to move forward efficiently in an AI first foot uh footprint you know in a posture and sometimes that's difficult with the

mechanisms you have in place if you have to bring in new teams new leadership move things sometimes that's what becomes difficult and I think that's what we're seeing is this reshifting plus there was a lot of hiring and a lot of positioning things that happened after coming out of covid uh that I think are being right sized now um but I also don't think this means that like those organizations are going to decide to do less or be any less ambitious. Even if they have smaller teams, even if they compartmentalize, even if they, you know, like re-shift, they're going to want to do more. And again, scale as that footprint grows and they get a better holding. And so, I just think that we're going to see a new type of job. Will some jobs be, you know, eliminated as we shift? Will some roles kind of evolve? Yes, of course. — Yeah, for sure. But also, there's going to be this new bevy of roles that we don't even know that like are jobs right now that people are doing that they're going to get paid for that just wasn't even feasible to think like, oh, you're getting paid for that. Like, how many people do you look at today and you're just like, wait, wait. You do what for a living? You know, — you put stuff in a spreadsheet, what? Come on. — Like, and so, yeah, I just think that in a few years the the umbrella that will encompass those things that you're just like, wait, what you do? You what's your job? Right? And I have a hard time explaining to people like my mom asked me, "What do you do? " I'm like, "Well, you know, like I do this thing and I build and then we're streaming on it. " She's like, "Why? " — I'm like, "Ah. " So like a lot of times — get a real job. — Yes. But imagine how — Yeah. Right. But imagine how many of those roles and things are evolving and places and — people that again you don't even know the problem they're solving and what types of jobs it's going to create along the way as they carve a new niche. Right? This is not a zero- sum game. Just because, you know, somebody builds a business and has jobs doesn't mean that somebody else can't also build a business and create jobs, you know. So, that again going back to like we're at the infancy of a new economy, a new way of building and forming. And we're seeing this in the community quite a bit. And so, — even with community, right, I've been in community for 15 years. Uh, and one of the things that I saw when I first began was when I told people as community manager, people are like, "Oh, you like are like the front desk person at like a, you know, apartment complex. " I'm like something like that, you know, but on the internet, let's just say, but at the time it just felt like it didn't feel real, right? And now we're seeing community become one of like the more, you know, core objectives that any, you know, large organization has because they know that in this new era, there's going to be so much more of a need to be able to connect with each other individually as people, you know, tackle our own problems. Even here on the chat, I'm seeing someone's like, "Hey, I have some issues with some of my apps. Like, I want to know how to build a little better. " and someone's like, "Oh, go check out community-help, you know, on Discord where you're going to find, you know, real people that have already built on Replet and be able to connect with each other. " And I think you're right, Rayar. I think that there's going to be just a variety of different roles. And I always think it's so wild when I see like the vibe coder roles go up where it's like we're just looking for someone to join our organization. These are, you know, large companies that are established and, you know, hundreds of of employees that are looking for people to be like the internal vibe coder. And you know, shout out to Ethan who was doing that at a previous role as well, now joined us here at Reply where he was just like people come to me, they tell me their ideas, I build kind of the framework and then I send it over to engineering to kind of, you know, bring that to life to whatever, you know, specs that need to be done within the organization. And that job would have been crazy, you know, 10 years ago. — And I think community is a really good example. And, you know, at the end of the day, like community, what is it? It's about forming connections. It's helping all of you get familiar with the product. It's supporting you. It's helping you build some connection with other folks that are also building as well. And it's just much easier to do more now with these tools than what you were able to do 5 years ago. Like 5 years ago, if you wanted to host a virtual hackathon and give you like the custom experience that we just recently gave you with the Replet 10 hackathon, that would have been really tough to do. The goal is the same. It's just now you have tooling to do more of that and to do it in a more interesting way. — Yeah. No, absolutely. I mean, I remember when Marcelo built out like a shell of what could be like a community hub. This is like very, very early, you know, three months ago. And it was something that it took us like, you know, it took us two years to do in a garage with Derk Anderson over at Startup Grind. It was unbelievable. That's went out to become Bevy. And it's just it's just — I don't know. It's just unbelievable. — Yeah. So, I think for like many teams like, you know, focus on what the end outcome is and then think about how to amplify that with this tooling. I think you're going to have like teams that are going to be able to punch above their weight and do that more efficiently and quickly, but I also think that just you're just going to be able to do more. I think Replet's a really good example as well. Like we are the poster child for AI, but we are growing like crazy. Like I think we doubled our headcount in the in the last year and so like we're all leveraging AI to the hilt, but we're also growing like crazy. And it's not and just we're just be able to do more and do it faster. I think that's the number one thing is like what is your ultimate goal is to create connections in the world around your product and your community. Okay, now how do we do that at a faster pace? it at a much larger level and then how do we like enrich the experience because

like now we have streams, now we have content online, now we have our hubs. Now you have a way to check out what you've been building the last couple of months. You have a nice profile that Rhyar is building. So end goal is the same. We want to make sure that you feel welcomed with us and you're empowered as a builder or like someone that's trying to take your vibe coding to the next level. It's just there's so many more ways for us to do that and to also make it fun and interesting. — Oh, absolutely. I mean, one of the things that [clears throat] uh Rhymer had built out uh is during the buildathon, he was able to see like, you know, usage uh like across like a wide board of like, you know, all the folks that were participating in it. And I remember at a previous organization I was at it, I asked for something like that for three years. It just it never came. How could you But it's like well you need to talk to data then data needs to talk to engineering then we talk about priority and it's like no we just push it out and for us it gives us all these insights into how to better help you right how to better support the community how to better talk to all of you even just a chat scraper right every single time you know every single one of these chats we're actually uh summarizing all of the things that people are dropping in so when you're dropping feature requests feedback or even just you know general topics that we're talking about all of this is in here and again thank you all so much for always dropping in engaging with us and a huge couple of shout outs here Ruth just joined us always Ruth thank you so much for joining Dino Kazir uh Ruchir as well big hello from India and uncle who's joining us as well so thank you always you know for being part of the community — yeah [clears throat] and I think it it's interesting here uh people are looking for closeness and connection I think is what this comment says and we're seeing things that they're referring to as the loneliness economy as people build right now which is absolutely kind of crazy to me let me switch back to this camera sorry I'm playing with the shots here. Just trying to keep the camera moving a little bit with our little Osmo here. Hey friends. Uh, a little behind the scene uh game review, they said. — But yeah, there's this um loneliness economy that's brewing, which is a shame to hear that this is like literally a thing people are seeing as a monetizable fraction of what's happening. And to me, this is a subset of what's going on with social media and the disconnect that's happening and how we connect. And also why I think community is becoming so important because people are looking for some kind of closeness, some kind of connection uh outside of their traditional like social structures and I think um a lot of those things are happening around shared interest, right? Things like creating, building things on the web, uh building apps, doing and again that's why getting together in a room is so powerful. Um, and so anyway, to the point of of what y'all were saying, I think this is only going to expand. And I think we're going to see again uh in a couple years how these relationships turn into fruitful, you know, revenue generating opportunities, businesses, like lots of things for people that they're not even thinking about now. Um, and also just, you know, maybe a little stability, maybe a little friendship. You never know, you know, who you're going to be along the way. — Absolutely. And I feel like a lot of the best communities, that's what happens, right? It's like, oh, how did you all meet? Well, I met on the, you know, X community. we were just really interested in building together and I mean previously I was over at notion and I you know still have many good friends from that community and people continue to hang out and not just hang out but also like you said build up businesses and continue to be able to thrive together and that's one of the things that I always kind of want to point out is um you know in the world of AI I think you know Ryar and many always talk about automating as much as you can and it's something that's really important but one of the things that's very core you know in our community team here is like relationships is actually connecting with all of you and that is something you cannot automate um I think it's one of the things that We try to like make it as easy as possible to connect with you as many of you as we can. But at the end of the day, like you know, that's why we have the Discord, live streams, and we always try to like, you know, take all of your feedback and into play. All of us here want to make sure that, you know, we're able to kind of uh, you know, help establish those connections. And it's been great to see. And I think you know the the right approach is that you want to use AI to automate all the drudgery, all of the tedious stuff that you have to do to free up time to do the stuff that really matters like forming real connections. Gazi mentioned this earlier. I asked him like okay what's the fine line of like leveraging AI too much and not enough? And his point was like hey like if something's tedious and mundane if someone wants to book an appointment that should be automated no problem. People want a response right away right? But if someone's getting a call from someone from Replet, you want that to be personal. You want that person to spend a lot of time with you. So if you're using AI to free up that time, you can do that. On our end, we got to do a bunch of stuff, right? That just like basics to like basic like maintenance stuff that we have to do in managing the community. And if we can leverage AI to automate that, we have more time to do in-person events and like spend more time with folks out in real life events. So I think that's like a really good kind of way of thinking about it. — Yeah. No. — Yeah. And I think that's a good way to just think about things generally, right? Like I think a lot of people, especially in this AI era, get caught up in the like I'm going to turn this like there's the meme where like the two people on either side of the email where it's like let's turn these three things into like a long paragraph and then the other side the other guy's like let's turn this long paragraph into like so it's like hey you expand on this and then the other guys do the summary like the AI to summarize it, right? And I think the metric you just talked about or the the funnel or filter

whatever you want to call it, that you just talked about is the right way to think about this is, you know, get rid of the stuff that's boring, that's tedious, whatever. I'm still waiting for the robots to like come and do my laundry and sweep the floors so I can focus on the fun stuff and make the art and do the playful things, right? And so, um, I also think this is a good piece of advice for some of you all as you build and as we talk about distribution and connecting with an audience is that you have to realize all these things are happening and you need to figure out how to break through that noise and how to connect with the audience or with the potential user with the whoever it is that like you're trying to reach. And that almost always comes with a genuine like we all want to feel listened to. heard. We all want to feel like we're part of a tribe, thing, like we're, you know, connected. And if your product doesn't do that, right? Like basically put me in the heroes role, right? And we talk about this a lot where like you get caught up talking about the product does this feature a xyz, you know? It's like, and you start describing the product rather than describing the user being successful in life because they have adopted your product, right? And it's this like uh Super Mario analogy, right? If you think about Super Mario, he's just a player, but when he eats this fireball flower or whatever, he can throw fireballs. And so, a lot of times, like we get caught describing the fireball plant, it's like, "Look at this beautiful plant. It's green. It's got these flowers, you know, blah eats it. " But nobody cares about the flower. All they care about is that when I have this thing, I can throw fire, right? And so you as a product developer, as a person who's shipping products, need to like put the the user in the perspective of Mario throwing fireballs, not talking about the flower that gets you there, right? And so I think again connecting with the real output, story and finding ways to like translate that into a way that you and I can understand quickly is the superpower here that you need to evolve. Um, and also trying to solve things that are closer to the ground floor, right? that are meaningful to people rather than again competing with the enterprise level SAS who's already got the market cornered on this big product idea that you think you're going to like rebuild overnight like that's not the way you know. — Um forecast number two or am I forecast two or three? I can't remember. Let's call it two. Forecast number two is I think you're going to see more creatives get into Vibe coding. We had Philip, our head of design yesterday on the show and he's just talking about how like more and more artsy types of folks, I call him the cool kids, the cool kids are getting into Vive Code. He had this really nice adage um something to the effect of code is the new creative medium — and you're just going to see I think it's just like folks are going to see it more as a tool to create something interesting and really cool. And up until now, I feel like, you know, the early adopters have been those very tech forward folks that are looking to build with AI. Then you have like the folks that are kind of the blend of the two. Like Rhymar is a great example of that. He's like a former designer. He's got a very artsy background, but he's also very technical, right? So he like Replet's like perfect for him. But I think more and more this year we're going to see like just outright creatives and artists start building in Replet. And me as like more of a technical person or more of a leftrain type of person, I don't think I'm going to be able to anticipate what that's going to look like, but I'm excited to see what they bring to the table and also add that to my toolkit as well because I'd love to be more right brain as well. — Yeah. No, absolutely. I mean, we see the amazing things that Kim part of our ambassador community is creating with like the replet movies. — Amazing, right? I'm like, I couldn't even imagine doing something like that. — No. And it was absolutely unbelievable because I remember I got on a call with him and he was like I'm not trying to create like uh you know animation. AI animation. It's like I'm trying to create a new medium, right? And that's I think that's where we're going, right? I think there's going to be so many more possibilities to be able to expand on the traditional sense of what's already happening. I think we were talking about that over dinner the other day, right? It's like there's this wide range and you can use it in different pieces. — Yeah, I've heard a couple people in the company talking about like the art of what's possible. Um, and I think that is like a everexpanding uh bucket that we can throw things into. But I wanted to call this out here. Somebody uh Wen uh Castiano say — is it just me or you guys look like sports commentators with years of experience defining the future of the game. — Wow. — Yeah, that's it. — It's high praise. Thank you. — Shout out. — And a quick plug for Vivecon that's coming up in June. That is the whole premise of Vivecon is what can creatives, what can arty people or what can cool kids do with Vibe coding with Replet. So definitely check out Vivecon happening in June. What's the exact date again? June 5th. — 16 17th and 18th. 17th and 18th. Yeah, June 17th. — So we'll drop a link to Vivecon. Head over to uh that site, get your tickets. Uh you're going to have to be there. Um, especially I think if you've been using Replet for a while. I think if you're in Reply, you're an early adopter. You're probably like me or Rhymar or Francisco. You've got this kind of leftbrain kind of mentality, very tactical. So, it's going to be really interesting for you to see what creative people, right folks can bring

to the table with Vive Code. And I think I'm going to be surprised. I know wild. — Yeah. No, it's going to be exciting. And we're going to have a couple different initiatives probably starting off next week for our community, for our folks that are in our Discord and, you know, here on the live stream. So, we'll have some little updates here on Vipon coming. — What's that mean? — We'll see. Tell me more. Francisco tease. Tune in. Probably Tuesday. I think by Tuesday, we'll have some — Francisco just be teasing. That's all. Can't just be teased. — I wish they brought this setup back to the office and did a few streams a week with it. It's legit, — man. You guys like the desk? Maybe we might have to do that. — Wait till you see what's coming, ladies and gentlemen. All right. So, let's see. You got any more on the list? What's up? predict? Well, I still haven't heard from the crowd what they think their predictions are. Yeah. Right. We haven't heard from you all in the mix. I see some folks in the comments and talking lots of stuff, but I don't see you sharing your thoughts. — Drop your predictions about Yeah. What do you think? — Where's the puck going here? It's just fun to do. Um, we touched this on this a little bit at the top of the show, but I'm expecting to see more agents being built out in Replet and more agents being used for folks. I think right now it's still very early for agent building and agent usage, but as Amjad mentioned, what Jason is doing here with Saster running this whole operation with two people and 30 agents, that's a glimpse of the future. So, I think next time this year, we're going to come back and there's going to be a lot more um live building of agents and then everybody's going to come with their own agents and look to optimize it. And so, I think more and more you're going to see agents and um I guess maybe it's helpful to think about like how would you think about the difference between a standard replet app or internal tool versus an agent? How would you make that distinction? So I think um defining an agent in my book is like um first of all it's just kind of like a fancy loop right um but in my mind the agent has context it has skills and it has the ability to kind of act on its own — whereas a replet app or any other software that you build unless it has some kind of deterministic workflow or some kind of like uh you know flow on the back end that's connected to AI you're kind of expecting the same results. So an agent you kind of have a probabilistic output whereas software you almost always expect a deterministic output. So when you're building something in replet, when you're building an app, when you're building a mobile app, a web app, whatever, you kind of want a very specific input out equal or input equals the same output with the same kind of sequence inside of it. An agent might make decisions on the fly based on different parameters, different situations. I think that line is being blurred. Yep. — Right. because you can now build in replet something that has AI access and it can do decision based things or it can kind of organize stuff in a way to like pull it in and then make decisions on that or adapt based on that new data. And so I think that line is blurring a little bit. — Um — but yeah, I don't know. I do see the ability to again create software that has a little bit of a little bit more intelligence in it. Um, and then once you can attach agents to those workflows, then that becomes like a whole new, you know, who knows what happens. — And one last thought, I'll add to this and then we'll welcome our guest, Jacob, to the stage as he's setting up. I think the way that I would think about, and maybe I need to spend a little more time refining this, but I would say an agent is a piece of software that you feel comfortable delegating work to. And so you can go and say like, okay, like our numbers are down for this group, for this conference. Like here's my goal. I want to increase this number by 100%. And like you go and research your context, go through a whole bunch of cycles and you use your tools, maybe even take some actions, start writing emails, and then report back to me when the task is done. So, it's a lot more like here's the task, go get it done, let me know if you have any questions, and then just let me know what the final result is. And so I think yeah, you're going to have more and more that you're gonna actually treat it like an maybe like as a like as an employee, as a partner that you can just push work to and then you just wait for the response back like you would with somebody on your team, right? Like here's the goal. Go get it done. Let me know if you have any questions. Let me know when it's all set. — That's what I mean. It's like that to me that's the definition of the agent is something that's able to then get that data back and kind of act autonomously without, you know, maybe with direction but with an objective in mind. And so it can kind of look and say, "Hey, do I have the outcome I want yet? " "No, not yet. I need to try again. Right? Whereas software is just going to run through the run once you're gonna have the output. And again, unless you feed that back into a loop, — you know, that's kind of what you get. So, um, that being said, uh, we just got a special guest on stage. What's up, Jacob in the house? Oh, let me do this. Hang on a second. You're — turn on his mic. — Oh, yeah. There it goes. — Oh, yeah. Now we got him. Good to see you guys. Good to be here finally on stream. You know, it's uh I feel like I see you guys streaming all the time. I never hop in. We've been

holding you for a very special occasion. You're our ace car, dude. When we got to get numbers up, get Jacob on here, dude. — Jacob, um, tell us a little bit about yourself. Tell us about what you're doing at Replet. — Yeah. No, I'm a field engineer here at Replet. Um, recently I've been doing a ton of design systems for organizations. agent scales for organizations. So, I sometimes I swear my entire job is just uh writing markdown files for various agents. um — the skills and stuff like that. — Yeah. You're like one of like the monster builders here at Replet, Dude, you're selling yourself short. So like when we talk to a new enterprise or a new business and they're like, "Hey, what can we do with Replet? " We throw Jacob in there and say like, "Hey, Jacob, show them the art of the possible and blow them away in terms of what they can build. " — Yeah. No, it's uh it's a super exciting time. — Yeah. Right there. — Oh yeah. — Yeah. There you go. Like ASMR I'm going to do. — Yeah. Right. There you go. — Um Yeah. So, one of the things we'll do with them is we'll take like their design systems, their component libraries, um we'll import those into Replet and have agent actually build with their design so it can match their company branding. — Um and actually in in certain cases use their literal front-end code when it's building, right? So, Replet's like this super agnostic harness um where it can work with any language and any framework and because of that it can use your design systems very well and kind of make this like really compliant uh front-end design — for your handoff to the engineers. — And why is that helpful for folks? So, I would imagine like one of our big um type of persona of users within an organization is a product manager or a designer and if you're a designer, you put together something interesting. If you're a product manager, you put together a spec and then you have to send it to engineering. Now you can leverage the your company's design system in Replet and just go prototype the thing. So I would imagine that's a big use case as well. Like you should empower those folks to just build. — Totally. Um I think like with AI in general, like it's super easy to get like a good-look design out of AI these days. You can even like some greatl looking ones out of AI, but like design is a lot more than that. It's actually really hard to do design well because you don't just want something that looks good. You would like how you want it to look. So specificity is a really unsolved problem in AI. — Um I'm sure you guys have seen that a lot before. You know, I just want this button right there and I want it to look like this, right? Um, so I think design systems are actually a thing that solves this problem because you know you've created this design system and it is exactly how you want it and because AI can use it, it can have that same taste and have much better specificity. — So I feel like everything I try and do these days with clients is solving for the specificity problem. — Um, where you're not just building it, you're building it how exactly how you want it, right? And that's just hard to do with natural language. Somebody mentioned in the chat that it seems like a really fun job. Would you consider it a fun job — sometimes except when you're like late at night 2 am like cranking on a demo. — It's it's a very like addictive job, you know. — Um I I'll like go home and watch some Family Guy and then uh just have some agents running, you know? — Yeah. — Um — we were talking about that last night. What was our distraction before parallel task while you like had the agent running and you're like waiting like obviously jumping between projects but maybe you're like on YouTube or you're playing like whatever and you're like just jumping back and forth between — Yeah. So it's definitely I was a software developer before this. Um I was very lazy software developer you know it's the lazy — the best ones — I was like yeah exactly when chatpt came out um I was like a miraculous I was so happy you know um and now all I do is write markdown files that's my programming language of choice now honestly it's like — what um so the the topic for today is what's next for VI coding is kind of the high level theme — so I know we're you know kind of throwing that at you without like much warning or anything but do you see any trends or any um future casting that you know like you might be um open to share uh with the with the audience. — So like if we're looking at AI in general um I think we're probably going to saturate a lot of benchmarks. We're real world use cases. So AI will become really good. And what's going to matter a lot less is the harness or which specific model you're using because they're going to be able to do everything that you want them to do so flawlessly, right? — Um so what really matters right now is the infrastructure. So you want you know your agents to actually impact the world. Like you look at like what is an agent, right? It's actually three core things. I'd probably say I'd say there's, you know, a system prompt or a prompt harness, right? That describes

how the agent should behave. There's an LLM underlying it, right? — Um, this the real engine. And then there's the tool calling and infrastructure built around it, right? So, I think like, you know, the prompt harness, all of this stuff is probably as optimal as it can be kind of across the marketplace. There's very small differences between harnesses. People often overindex on that. Um, but then there's the LLMs, right? And those are relatively similar, at least in the same ballpark. Um, but what really matters now, and I think what where there's a lot of alpha, uh, is to have like really good infrastructure, — right, and really good tools. — So, the way you interact with the world is becomes super important. like an agent that like a marketing agent that has access to email is going to perform obviously way better than one that does not, right? — One that has better access to write more functions in email or whatever it may be um is going to still perform substantially better. So, it's all about the infrastructure. tooling capable and that that's actually like what we do really well at Replet. Like even internally, we have internal tools that spin up our development environment so agents can work on them, right? Um, that's kind of like our our bread and butter here, Replet, you know, spinning up a development environment so an agent can work on it. — How much of that infrastructure is coming from Replet versus the infrastructure that individual users are building on Replet? I guess is that just like relying on Replet to like increase the performance of the build and the platform or is there like something that you're seeing users do to like help optimize that? — Yeah. Well, there's two layers. So when you're building software using like replet agent that's using the replet infrastructure right which is very agnostic and we spend a ton of we spent around 10 years building that infrastructure right — um when you're building agents you know um to do various things. Uh you're going to need to configure these tools and this infrastructure around it and you're going to need to think about like what is the best way for my agent to interact with the world. So on Reply we try to make that easy u by adding the all these integrations, all of these different ways of adding infrastructure and tooling to your agents. Um, but a lot of it is just going to be thinking about like having really good design intuition or systems design intuition — um — for like how should my agent interact with the world and being opinionated and being uh concise with how you present these tools to agent and benchmarking uh how it performs is super important to that. Um, so — yeah, this is and this is where I I um I agree with everything, — but I also think this is a good time to point out like how early we are in some of this, right? — I made the comment the other day that just the fact that we have to like write all these skills files and that we have to kind of hold the agents hands is just an indicator of how early we are in the space. If you think of like the old punch card days and the internet format like when computers first started and you know I kind of had this manual process of switching between programs and like even remember like I maybe some of y'all are too young for this but remember uh playing a computer game and you would have to switch to the next disc to like advance in the game right and so I think we're kind of at that phase of AI where it's like context windows has some limitations the you know the scope of it's not learning as fast as it could and you still have to kind of write and give it these rail Yes, it als certain things because you're human. You've experienced the world and we kind of understand how things work. And so I think you know like again this is just how early we are that we have to manually hold the hand through some of this stuff. I think over the next handful of years we'll start seeing that stuff evolve and maybe we'll then have to move to doing other parts of the process to keep it properly structured. But like I think some of the things we're doing now are a little primitive and a little like we're going to look back on and be like, "Oh, remember when we had to like write MD files, you know? " — Yeah. Cuz there's like a bunch of decisions you have to make to optimize your agent. Like we were just talking to somebody yesterday and they're like, "Hey, how to get all of my context into agent? Do I just drop in PDFs? Do I extract it from an API? " So you've got like how are you going to provide act context to your agent? You also have the question of what tools you're going to provide to your agent because you don't want there's an optimal amount of tools that you want to provide, right? not too few so that it's not helpful but also not too many so it doesn't get confused and also the instructions and the skills. So there's still like some things that you have to do as you're customizing that agent experience. I — was trying to think of Jacob at his uh Thanksgiving Day table 20 years from now. You know kids, we had to write the markdown files — uphill both ways. — They had these things called skills and they were like in this really weird format markdown. Crazy kids. — Yeah. It's when uh agents weren't smart enough to have skills of their own. [gasps] — But I want to go back to something that you mentioned earlier around the specificity and that like that's really

where the off is and I think that's spot on because say like two years ago you're a designer, you got a nice UI and looks beautiful, but when that design goes over to the engineering team and they actually have to build out the backend and the actual UX of the application, things just would break. So it's much more straightforward. you as a designer, you build out that prototype so you can click through different things and see what the like the back end how it's working and then like have that be the final thing that you send to engineering. — Yeah. Now is a now is an exciting time to be a designer I'd say. — Oh my gosh. So much advantage they have. — Yeah. That like you have this like component library where all of your design and core branding lives, right? And a lot of that tends to live in like a Figma right now. Um and a lot of it should really should. Um, but what happens is like these front end engineers will kind of lag behind or things don't match the Figma spec. So I see designers starting to code a lot more, right? So they'll take those Figma designs, import them and actually build the code and then hand it off to developers to actually build real UIs with it instead of just like, oh, this is two pixels off. Please fix developer. — Well, and that that's I mean, as somebody who's designoriented, design-minded, like I've been on the other side of that for a long time. I'm like I had the idea for the product. I had this idea for what it should look like and how it should interact. But then that translation phase of trying to get that out of my head to a developer's head. Even if you give them the most beautiful nuance, like you said, specific designs, you might have some idea for how this transition should happen or how things should flow from one state to the next. Or again, when you're in Figma, when you're working in a canvas or you're doing something like that, it's like this looks perfect at this screen dimension. However, the screen's going to do this and this and this and this. And so like the designer wants to have some kind of input on like what happens in each of those phases. Like how does this expand? How does it respond? How does it do the thing? And a developer might have to guess and not have the same aesthetic or the same context or the same idea as the product ideator or originator. And so like now we're in this phase where the designer can be the person that gets in and makes those nuance edits or defines the responsiveness or just says, "Hey, maybe when we go to mobile, I want this to be on the top instead of this, which is the natural flow of the HTML or the CSS. " And so all of those things are decisions that you either had to again get real specific in the brief or be like some unicorn that had the skills to design and develop the thing yourself, which almost never happens. — Yeah, I can tell you felt this pain. — Oh, this is the pain. — I was like, I want to hit the bullhorn. That was good. Oh, wrong one. — I was like, good one. Clip that, dude. You can tell you've been on the other side, dude. That's good. Oh, yeah. No. Yeah. Know, they should be owners end to end, you know. It's finally It's so exciting to see that happening. I think one of the coolest things I've seen a designer do on Replet was they built like a component library but for motion like a motion library — and they had it all tweakable and they said here are all the bezier curves that the all motion happens right and then they just had click copy paste as an agent skill — and then they paste it into whatever cloud code or replet agent skills um and then all the developers had the standard uni unified but it was just this one designer who had this entire like motion library which I thought is super cool. You can't really do that — before now. — Yeah. — How are people getting their Figma designs into Replet? I know we have an MCP. You can just drop those files and images into the prompt like what's the best way to do that? — Yeah. No, honestly, um if you just hop into Figma, you can go to whatever frame you like, you can uh right click on and copy paste the link to it and then just paste it into the agent chat and just say uh make a copy of this. Exactly. Make no mistakes and then it'll do it. — Make it awesome. Make it beautiful. — No mistakes. — Wow. — That's amazing. — Yeah. Exactly. — I think we got to um do a guest swap a little bit. Yeah. I think we got I've got the next uh we got the next guest coming up. — Good being on, guys. Oh, yeah. We're going to go to this first. — Heat. — What's up everybody? Welcome back. Uh we were just talking with Jacob. Today we got Samuel, who the crowd has met before with one of our more popular videos about videos and all sorts of stuff. Yeah. And uh if you're on X, you've seen them all over X with different product

announcements and different things. Uh really the slides champ. Like you're like the OG and killer of like viral slides videos and stuff. So anyway, Samuel, welcome to the stage here at Saster. Yeah. And just to set the scene a little bit, we're talking about um what's next for Vibe Code. So day one, we kind of uh reviewed a little bit about like what the state of the market. We had Amjad, Michael and Gazi kind of, you know, sharing a little bit about where we are now and kind of how they're approaching the market. Uh yesterday we talked a little bit about how to thrive in the space and how people are making money and kind of where the tooling is today or uh now and today we're talking a little bit about like what comes next in this space, how do things evolve. Um and so anyway, just to frame you in for the conversation we're having here with everyone today. But yeah, maybe just say hi and introduce yourself a little bit to the crowd. Yeah, great to see everybody or meet everybody if it's the first time that you're seeing me on this live stream. Uh I'm Samuel. Uh I'm on the product team at Replet. Uh I joined about a year ago. Uh before this, I was a repeat ventureback founder. Uh raised millions of dollars, scaled businesses to hundreds of thousands of customers, millions of dollars in profit. Um and yeah, super pumped to be with you all today. At Replet, I've helped create the products Replet slides, Replet animation, and then also working on our canvas, our core agent launches, uh and a lot of other stuff in between. — Yeah. underelling himself a little bit there. But yeah, that's a good start. — Yeah, absolutely. So Samuel, we're here SAS or 2027, 12 months from today. What do you expect to be different 12 months from today? [snorts] — Oh man, I mean look, I think uh Replet and this space is so interesting because I think uh so much of what happens in the you know app generation, code generation space is downstream from what happens at the model layer. uh you know I sit with the our AI team at replet doing applied research and you know it feels like every week or every two weeks uh some you know new model comes out that has you know maybe like one piece of capability that exceeds what was possible before uh and then sometimes that unlocks like really new product functionality that's how replet animation and slides came about is you know we got to play around with the Gemini 3 model for the first time back in November uh which led to our design mode launch and uh I just I saw this capability ility uh that came out where okay we could actually produce better slides than anything available on the market and these like motion graphic videos with code that weren't possible before you needed After Effects and a human designer for and so yeah if I forecast out 12 months I assume that code generation will get better the models will get smarter uh I think we're on a continuum uh towards postp prompting I think that's like the current era so instead of just uh you know you as a human having to give super detailed instructions of exactly what uh the agent should Now it's becoming easier for the agents to sort of know what to do itself. Uh and so at RAP we uh both uh you see this in like you can prompt the agent in a simpler way one and then two now with our new follow-up task feature where after you uh finish a task, the agent will actually proactively say to you, "Hey, here are some things I could do next. " And then you can just kind of click to accept them and it'll do them. So I think that'll just continue to uh happen and agents will get more and more autonomous. — That's spot on. Like if you compare the like how prompt engineering was so hot a year ago like these crazy prompts that people were doing and compared to what I have to do now like one sentence like build me this amazing slide on this topic make it look awesome like I think that's part of that you're referring to just going to be easier and easier for reply to do what it needs to do. — Totally. Yeah. And I think part of what's happening that maybe the average uh person doesn't realize is that these model companies, whether it's OpenAI or Anthropic or Google or whoever else, uh they're doing a ton of post-training on these models that are making it so you don't have to be as specific when you're prompting them. So what do I mean by that? Um you know, you have a pre-training phase where you take a bunch of data and then the model learns to predict the next token. But then there's also now a post-training phase which is increasingly a focus for these labs where they say for a task like you know generate a pretty website they'll have a bunch of uh you know sort of evaluations of you know is this website pretty? Maybe human experts uh play a role in creating these evaluations. Um and then the model will go and generate a bunch of websites and it'll get graded. Is it pretty? Yes. If yes, reward. If no, don't reward. And then the models, you know, inherently get better at turning like a one-s sentence prompt into a beautiful website. Whereas before you would have had to copy paste like you said, you know, 2,000 words from some Twitter post. I was talking about this on stage yesterday at Zaster, but you don't need to be an expert at prompting to get great results from AI anymore. It was pretty cool. — Yeah. — Remember when the prompt libraries were a thing and the prompt optimizer like that was all the rage like a year ago? Like here's my prompt library. Like that's no long. Um we talked about like what's going to be different and some of that's going to be driven by LLMs and how much better they get at things, but also I think it's on the community side, right? As we release some of these products out into the wild, we see what people are building and we're like, whoa, like I didn't think you could do that. Let's like try to build that back into the product.

— Totally. I mean, look, when we uh released light animation, I just intended it to be a solution to a problem I had as a founder, which is like, hey, you know, I built this thing. I need to tell people about it, and now I'm constrained by like making the marketing video to go and share uh this new product with the world. We actually had that problem at Replet. We were doing a launch last minute for this product called Fast Mode back in November. Uh we didn't have a launch video. That's why we built Replet animation. But once we released it, we saw people doing all sorts of stuff that I did not expect with the product. Someone made like a full feature film uh with Replet using I was like, whoa, this is crazy. Um and so yeah, I mean we just see this across every single product category. Like our community is amazing. Uh and as soon as we release something interesting, they find, you know, 100 new things you can do with it. And it's only going to happen more and more. I mean, we talk about — I talk specifically about almost anytime I get a creative problem or somebody asks me for some kind of task or anything, I'm like, — can rep can replet solve this? Let me see. You know, and then it's like, okay, maybe not on the first pass, but maybe on the second pass, maybe I got to refine the prompt. And so it's like, and more the solution is yes, right? More and more the answer like, okay, cool. You know, like we can do that. And so that's going to be interesting to me again like how do people start pushing the boundaries of this thing in ways that we haven't as a company thought about like oh they're doing that with it now hm you know it's like those are really cool moments I think and I think um the culture of the company to allow those explorations to happen to kind of create opportunities for those things to take place. Uh we've been talking about there's been another theme over the last couple days which is the dog fooding aspect of you know internally we're using this tool not just to like build a tool and sell to the world but we're like actually using this actively like we are probably the first and foremost users of this product internally. And so that shapes a lot of again the needs that we have or the little slivers of things that we find. And so, um, maybe you can give some insights on that process or, you know, how that stuff happens on the back of the, uh, behind the scenes because I think a lot of people would be interested to know, you know. — Totally. Yeah. And look, and, uh, I'll be super open and honest here. Like, when I first joined Replet, I was like, you know, I'm, you know, decently technical. I love cursor. You know, Replet's a cool company. I've admired for a long time, but like the power of Replet is to take someone who's super non-technical and give them AI. But Replet's not really a tool for me was my attitude. Um, and that's just really changed uh over my time at Rep, which is so cool to see like as me and the rest of the team work on the product and we listen to the community, we've made the product so much better, that like when I first joined, it did feel, you know, there were definitely people that were using Replet a lot at Replet, but there were also people where it felt kind of like pulling teeth to be like, "Hey, come on, like use the product more. Figure out how this can help you. " And then with the launch of design mode, I just instantly proliferated across the entire staff because they were like, "Whoa, I've never seen AI do this before. " And for me and the product team, we started doing, you know, a ton of our prototyping in Replet directly. Whereas before, maybe it was, you know, occasionally easier to use Figma. If you knew Figma, uh, Replet became the best tool for the job. And now I've really seen that happen with everything else. Um, and I find myself rarely reaching for — I'd be interested to know like I want to pull on that thread a little bit just kind of personally because like I only know Replet — coming from the user side, right? Like I only know it as like this tool that started to unlock me creatively like to build and write code. And there were for sure times I think I started using — early uh right before agent 2. So I was like just tinkering with agent one. I was like this is kind of something here, you know, like you can spit out a little something something. And then agent 2 came around and I was like, "Oh, wait. I can, you know, and I started I built like a little membership site and I remember I started tinking tinkering and that's when I first started kind of posting about like, hey guys, there might be something actually here, you know? " And then agent 3 comes out and it was a little ambitious, right? I was like still I stayed on agent two for a while cuz I was like I'm kind of comfortable here. I can keep this thing locked in. Agent two or agent three kind of like but then the models caught up — to the harness and the things that were happening and then I was like, — "Oh yeah, — this is crazy. And that's [clears throat] maybe what is that now 6 8 months ago something like that and that was really like a moment for me and so like I only see Replet from that side of it because I was not technical. So I'd be interested to just pull on that thread a little bit more about how that transition was for you. Um maybe how you still use it now or how some of the internal team uses it now. Is it mostly just kind of testing or like you're saying it's becoming more and more just part of your daily flow? — It's becoming part of our daily flow. I mean we even have an internal AI uh like innovation team uh that is tasked with taking replet uh and AI in general and figuring out how can we speed up our workforce and so Luca and SI like they just go across the whole org and they find all these inefficiencies manual processes and they automate them with rep it's insane like sometimes

they're spinning up like hundreds of tasks to go in one project uh but I mean it's really helped our support team our recruiting team um And so it's amazing to watch what they do in terms of internal tooling. And then for me uh I do play some with our internal tools, but I think the daily driver for me as a product person is definitely like prototyping. Uh as well as replet animation for making marketing videos. I mean I I've made at this point like 30 plus posted videos with replet animation that have collectively gotten like more than 10 million organic views for the company which has been awesome. Um, and then yeah, dayto day, like literally this morning on my car ride over here on my phone, I'm sending these prompts to the replet agent to work on this prototype to present to the whole team uh tomorrow to talk about the future of one of our products. And so it is really like a core loadbearing system for everybody at Replet. — I I've chatted with a couple of technical folks that have come to use Replid and there's two things that stand out. Number one is things that you can only get in replet like design for example you mentioned but it's also that you know even if they're technical they don't want to do some of the more mundane DevOps things that are related to building something new you don't want to have to stand up off set up a database you don't have to serve it somewhere host it you just if you have an idea you just want to quickly put it into replet and just deploy it right there all in one place in replet and so for even if you're a technical user you don't want to do it with especially for like a side project or like you know again like a zero to1 dash board. Even like a repeat venturebacked uh founder friend of mine that I was catching up with last week, you know, his current company's raised like 10 million plus dollars. He's made a ton of money in the past. Very technical. He still uses Replet today. Um he has like his core, you know, production code base. He he's built himself separate from Replet, but then all of these sort of like one-off dashboards, tools, — things that are still like userfacing. Um — that's what we tell people all the time. It's that like rapid iteration. if you're tinkering with stuff, it's like there may be a point. Somebody asked us the other day like what would be the benefit of hosting offer epitlet versus not offer up epitlet and our answer was like — basically nothing unless there's a certain level of scale or um maybe like customization you want and you have somebody very technical and you kind of want to own that. But even then, you know, I'm seeing more and more technical folks that are like, "Hey, for my experiments, for these little quick things, especially if I'm just going to spin something up today and I want it out tomorrow and I don't want to think about that stuff and I want it to autoscale and you know, there's lots of things that it just makes sense. It just works like especially if you've never like programmed before. Uh, Replet is like the intuitive way these things should work. Uh, but this is not how things actually work. " There's a lot of work that goes behind the scenes at the experience feel as easy and straightforward as it does. — And I also [clears throat] have to mention that as the lines between the technical teams and the nontechnical teams blur, you're going to want to have one place where everybody can collaborate. So if you're like doing everything locally, you're doing things via GitHub, I would imagine it's going to be hard for you to pull in that marketing person, that design person and say like give me like valuable input on this particular prototype. But if you have it in Replet, you just send them a link, they jump in there, they can ask questions, they can kick up stata tasks, you can review it, and you can merge it back. So if we're moving to a more collaborative type of experience, Rep is going to be really handy for that. — Yeah, I mean, maybe this is a good time to talk at least briefly about parallel agents. If you haven't already, uh go a hot topic. — Yeah, go on Twitter and go look up uh parallel agents on the Replet account. We just did a launch video uh at the start of this week that I helped work on that kind of explains why it works and why it's important, but I'll try to do so briefly here as well. Um replet parallel agents is really the first time uh that AIS collaborate like human software engineers. So what does that mean? Um you know sub agents have existed for quite a while. So that's the idea you have a main agent uh that's working on you know single instance of your codebase. It might spawn other agents that then are, you know, working on different parts, but it's all happening in the same copy of the codebase. Why is that a problem? Well, usually that's happening locally, which means I can't invite my colleagues to work on things at the same time. It also means that like everybody's stepping on each other's toes, right? One sub agent might be editing the file system as the same time as another sub agent. They're not really sharing context properly. And so, what did we launch with parallel agents? Well, we basically launched the ability for, like I said, agents to work like human software engineers. How do human software engineers work? There's GitHub up here with a copy with the main version of the codebase. Everybody has their local computers and they they'll clone the main copy of the codebase. They'll work on a branch and then they'll push up just their changes to GitHub and that way people avoid stepping on each other's toes. This is how agents now work in Replet. Uh and so uh there's sort of a main agent on your main thread, but then at any time you can spin up a task. Each of those tasks copies the codebase and then actually works in an isolated virtual machine. Meaning it has its own compute infrastructure. It's able to run the codebase end to end itself. It's able to test itself in this isolated environment and then they'll merge their stuff back together agentically. So maybe if you're not technical, you don't understand how complex and crazy this is. But uh

yeah, it like replaces a team of software engineers and it allows like me and Manny and Rayar and whoever else to all be working on tasks in parallel in the same ripple. We were just talking about this at dinner yesterday that we need to do some educational content on like what is actually happening and the git and the merge and the workflows and helping people understand not necessarily to know or understand like how or why but what's actually happening back there and how this is is possible and kind of just letting people know like hey these are the traditional workflows this fork create a new branch make some edits merge it back to main like this is all now kind of happening behind the scenes and the agent handling and it is a big deal. It's like it seems simple. We're showing the video now on screen for folks who are watching. Um but yeah, it seems like a simple concept, right? Like hey — you work and you somebody asked me like how does that work? It's like well have you ever been in Google Docs and somebody's like you know editing at the same time and you're working on this paragraph and I'm editing right behind you and we're all kind of in the same project. Uh it's obviously not the same exact kind of concept, but it is the in the sense that like we're working on the same document without each of us having to manage a separate doc and then at the end figure it all out like the agents kind of doing the figuring all out all those changes and how they they merge back together. So — yeah, I think the term is merge conflict and this has been like a topic that's plagued teams like forever in software engineering. — I'm dealing with it right now. Literally, I'm sitting right over there and I pinged my colleague. I'm like, "Please help me fix this for one of my branches. I need this fixed. " — Oh, that's awesome. And then I It's only possible because we like this is like from off of a fiveyear infrastructure bet, right? This is why we're able to do this. Like this is why it's going to be hard for other people to do this. — Yeah. It's nuts. I mean, I think it's really only us and then Cognition, which is a tool for like uh full-time software engineers at big enterprises uh that have this functionality. Like that's it like and the whole market. Uh it's really hard thing to build. — And not only is it's a hard thing to build, build unless you go back five years and make decisions that like allow you to be here and make that decision. Yeah. Right. Uh so — let's make it easy. — Yeah. Um so and I think we talk about this a little bit um in other ways too that um there are things that like folks claim or will be able to claim that they do a lot of the same things. Uh but as you learn more and more of the nuance of what really separates the platform, you're going to start understanding the some of those complexities and uh maybe you can talk a little bit too some of those infrastructure bets or just generally like the footing we're on compared to to other folks who are in the space. — On the one hand, it is straightforward to build a tool that generates code and shows you the results, right? uh like in replet I built in you know we launched this sort of like faster agent in December that was able to you know oneshot web apps in like 3 minutes and in like roughly 3 minutes I built an AI slide generator in replet that you know you could type in a prompt this is not like AI slides in replet this is like an app in replet that then you could sell to generate AI slides and I oneshotted it but you know it wasn't very good right uh it did not have you know security in mind it didn't have like native you know author database on behalf of our users. Um there there's just like so many different products that go into our cloud offering, our security offering, um to be able to support uh you know, loadbearing infrastructure for hundreds of thousands of users on your apps. Um and so while it's easier than ever to build something that generates code, uh it's more competitive than ever to build like a comprehensive platform. And I think that's the reason why you'll see I mean people even on other vibe coding tools will switch to replet once they want to like really launch and achieve scale. uh because it is pretty hard to replicate like the breath and depth of things that we have. — And now we make it easy for you to import those projects that you maybe started elsewhere into Replet. Check out our free importers. Quick plug. Um can you maybe give us a little insight in terms of how you see the Replet experience changing cuz like you know now we've got slides, animations, people are talking about agents. So like where do you think like that's going to like how is that going to evolve on our platform? What people can expect? Yeah, I mean look, it's all just code. It's just code. Uh different agents generating code to produce some end result. That's how I think about it. That's why while it might seem that like slides or animation could be a distraction or like a big product investment versus our core agent pad, it's really not. It's just code. Uh and so I think replet will continue to be, you know, better at generating code to do any task. And that's true whether you're like generating uh an end like asset like a you know whatever marketing uh ad web app for your users anything of this variety or even just directly doing work for you right can update your linear tickets or pull information from Slack um you know sort of being a digital co-worker for you is also just like a product of generating code and I think we'll just continue to make that easier and more and more

autonomous — you expect the same interface like you come in you start with the chat or do you like expect that change? — I don't know. I mean, again, it's all it's all so dependent on the — flow, right? — Yeah. I mean, uh, expect, you know, who knows? Maybe it's wired to your brain in two years. Well, we'll see. — Some neural link or neurolink adapter connector. — Well, again, hey, we had a question here, Sam, that you might be able to answer from the crowd. And I'll take a little bit of a stab here, but will they ever be able to use Replet offline? — Um, and I — Yeah, go ahead. I'm going to guess probably not because Replet deals with a lot of like online primitives. It's an onlinebased uh browser and when you're trying to preview and access a lot of the stuff you kind of need like some of the primitives of the internet. — I think the big thing you could write, right? So I mean it's if you're running like a trillion parameter model, it's very comput intensive to serve that inference. And so you know you can't run these frontier models like on your MacBook. Uh now if you have like a you know whatever $50,000 like Nvidia GPU at your house uh and you're running you know whatever Kimmy or some open source model you know would it be theoretically possible to vibe code like locally at your house like maybe uh whether or not that's something we support you I don't know. — Yeah. — Well then what happens like when you need to preview something that's again attached to a web service or attached to some thing or has some dependency that's connected to the internet. So it's like we're literally building services that are alive by default, which means they need internet connectivity to like do the thing. — Yeah. I think I would push back and say like where do you need to run it offline? Like I even had a buddy of mine, he's an engineer and we were debating. I was like, "Oh my gosh, like Amjad was so right about cloud-based development because even having like local agents versus cloud agents, cloud agents are clearly going to win because I can collaborate, you know, I can hand off between devices. " And my buddy was like, "Oh, but like what if I'm on an airplane and I just want to code locally? " I'm like, "And not have AI? Like come on. " Like — get yourself a Starlink covers everything. — Exactly. You're going to need the internet, I feel like, uh to do work. — And I think this is a hold over from like early developer days where you're just like you're in tune with the code. Your code only lives on your machine. And when you run it, then maybe you simulate an online environment. So you can run your code in a live online environment, but the developer themselves is kind of in full uh yeah uh full uh ownership of that code and kind of what's happening inside of it and they could work offline in that instance. Um but again like you're saying like it just seems unlikely, incredibly unlikely until the models get way advanced and maybe you have a big server at your house but then there's going to be other reasons you're going to need the — very neat. Well we know we got a hard stop here at the 30-minute mark. Thank you so much Sammy for joining us. Always a favor here in the chat. Thank you so much. Forward to what you push next. — Good to see you all. — We'll be right back. — Thank you. — Oh, oh, here we go. All right, so we got some uh comments in the chat. We did a little flippitydo and we got Kathy over here now instead of Samuel, right? We get a little We go to commercial break, come back, we got a new guest. That's how it works. We had to close out with Kathy. She is one of the favorites at Replet and I think she's a favorite with all of you out there. If you don't know Kathy, you probably already interacted with her with the Replet support handle on X, on Reddit, on LinkedIn. She's all over the place. She's my favorite friend on X when I get those questions and those issues around people having issues. Like I just tag Replet support. I know Kathy's going to be on it. — Yeah. No, absolutely. And one of the things I really want to shout out is when I first joined Replet, it was a different time. uh you know this is July of last year and uh you know we didn't have too much support uh you know the company's going through a transition from you know the previous kind of uh focus to you know where we are today and uh truly when like the moment Kathy joined the team I immediately felt it the community felt it and um you've probably interacted with her you just didn't know it um she has an incredible voice on Reddit um also uh has tons of experience on Discord and helped us kind of navigate uh you know our Discord today and um just overall support. She's been massive. If you've ever sent me a ticket or any one of us a ticket, Kathy's probably the one that's jumped onto it. So, — save us, please. — Truly, truly honored to have her um on the show. Thank you so much, Kathy. — Oh, you guys are making me tear up. Thank you so much for having me and good to see you all. Yes, I'm on social media and I help out with customers, support you all. I'm a bit of a gamer, so I'm going to throw some links. — A budding streamer yourself, right? — Yeah. Yeah, I'm also a Twitch streamer, but uh going back to the game reference, I feel like if you play a lot of um Overwatch or Marvel Rivals, for example

these hero shooters, I'm like the heels. I help out all the engineers who are like DPS and you know, everyone else on product and marketing, they're like tanks, DPS, and I'm the heels here. — Oh, what's Rhymar? — Um — is he a tank? — I think you're DPS. I don't know. What does that say? — That means you output the most damage. You're like a damage hitter, bro. — Hit or bad? I don't know. — Well, it's good. — Okay. How about me and Francisco? What What's our player type? — So, I think you're also DPS, baby. Is a tank. — I like that. — Bring the bong. — Yeah, it's so amazing to have you here, Kathy. Thank you so much for — Oh, they're commenting on your nails. Thank you. The riplet orange. — Hey, nice. — That one's broken, but I had it orange for our 10th birthday. — Ah, nice. — Orange isn't really my color, but then ever since I joined the company, I see it everywhere. So, like I'm getting used to it. — Your favorite now. — Yeah. — What do you stream? Tell us about that. I'm interested in what you're doing like when you're not on Replet support. — Yeah, I play a lot of Starcraft 2. That's my main game. So, it's like a real time strategy game. Um, I used to do it fulltime, but now it's more of a hobby. And now that I'm at Replet, I'm like hoping to vibe code some stuff to put into my stream. So, you guys, I've already talked to you guys about the whole live streaming experience. Um, but yeah, I made my own landing page as well with Replit. So, that's my start for now and we'll see how things go. Well, we have this thesis that vibe coding is going to get to the place where gaming is right now where like people will tune in and watch somebody play video games. We think that there's going to be that same type of demand for people tuning in and watching somebody vibe code. We see that with Rhyar's deep dive build like he's he goes three hours, starts with a prompt and won't go offline until that is that build is complete. And we see a ton of people loving that content. So maybe you can tell us a little bit about that. What do you think Kathy? think that vibe coding is going to be as big of a deal as gaming one day? — Um, I think so. I don't really see a lot of live sessions. I think most people do short form content or they have curated VODs or videos, but I think it would be cool to live stream just to kind of share that experience. Um, I'm really impressed that you guys can do it on the fly. Uh, when I'm vibe coding, I'm like I'm kind of waiting for my agent to be done. So during the downtime, you guys have to, you know, fill up the empty space with conversation, all just kind of waiting for things to be done. So that's pretty impressive. — It makes for good TV seeing Rhymar sweat and debug stuff. — Yes. — Well, there is. We were talking about this yesterday that I see a lot of parallels in the early kind of esports formation uh to what's happening in the content space as far as like vibe coding and this AI stuff where uh people are experimenting a lot in different media uh realms and I think over the next handful years we will see like some folks kind of become standout players in the space just like we have I don't want to say household names but if you're a gamer there are like lots of names that are prolific. Uh each game has its own kind of vertical of people that are really good at it. Um e even like sim games versus like roleplayer games versus like first-person shooter game like they all kind of have their own ecosystems and their own like worlds that you can kind of dive into and people that are uh playing and streaming and blah whatever. And so I really see this also following similar paths. Um, and I see a lot of the same, we talked about same dopamine payoffs and also when we're doing those streams, right, there's almost this like uh tension between the audience and like what we're doing. It's like, — will he pull it out? — They might fail, right? — Yeah. So, there's this like there's this excitement of like — ship. That's good. — We might watch these guys fall on their face today, you know? — Yeah. I think that's a good point. Um, before there were live streaming sessions of people co-working, they're programming. So, I imagine vibe coding would be um, very easy for folks to follow, especially if you're using natural language, right? — Sometimes I stop by the programming streams and I'm just like, I have no idea what's going on. I'm not technical, but it looks cool. And I like the vibes, too. I'm usually there more for the person and they're chatting and the community if there's other people talking in the chat and whatnot. — Yeah. And I think what it helps people understand at least like the feedback we've gotten um so far has been that it helps people kind of understand how to approach the agent. Sometimes it can feel a little difficult of how am I supposed to start this project? How do I deal with this? I don't know programming. I don't know what to ask for. And so really when they see me hit the function key and just start talking, you know, to the agent and they're like, — "Oh, you you're just talking to it. " And then they're like, "I could do this. " Right? And so like part of the streaming process and I think like it's very easy to do those polished demos or to record something and you then you do 10 takes and then it's like perfect and then like maybe you tried the prompt 50 times and the one time you get the

output and you're like oh this is the perfect thing right and those make great marketing videos they make great Tik Toks they make great like wow look at this thing but what misses as part of that is like the reality that most people are going to go experience when they get into the tool and they have to try a bunch of times like you know change their approach a little bit. Um, and I think that, you know, the live streaming or this type of content really humanizes that, really like shows like, oh, okay, uh, these are the pros in this and they still struggle with X, Y, or Z. And so, it really just shows people like, oh, this is approachable. This is something I can do. Um, and it kind of gives them some insights into how to get started. So, uh, again, I think that's like that's some of the benefit of this type of content. — Yeah, I think um, Vive Coder streamers can inspire confidence. uh people can show up and they'll learn from you live. — So, it's also great educational content because I've seen some of your live streams are very beginner friendly, kind of walking through the steps, explaining the terms. So, I think it's a really good introduction for people [clears throat] learning how to buy code. — Kathy, I wonder if you can tell us what are like the more most common things that you see people having trouble with on Replet, but are also like the easiest to fix. like what what tickets do you see coming through like, "Oh, that's such an easy fix. Just do that. " — Um, I think first and foremost, I want to say prompting is a skill. — So, I think most folks, I mean, we all think we're good with our words. We have our egos. But if something doesn't work, — yeah, maybe try a different step. Um, you have to kind of coax the agent sometimes. — I know. I yell at the agent. I don't curse at the agent. I don't say firm with agent. — I do yell. I put in caps locks. — And I learned recently that might not be helpful. Um I was interacting with my chat GPT and I just kept yelling. I didn't curse, but was like, "It's not working. What are you doing? " And it started crying. It started using crying emoji. Really? Oh my god. Oh, I need to start a new chat session. But it's like it's not helpful. So, um I think just kind of working with the agent like, "Okay, let's approach this from a different angle. this isn't working right now. Let's try some different steps. Just kind of have a new way to prompt. — Yeah, — I see on Reddit a lot of times or just in general support that people are like, "Hey, this is really expensive or how can I how can I, you know, uh save some costs on this in general? " And you always have great answers like anything you can like summarize for our audience for folks that are learning like what are the most efficient ways to use the agent? — Got to write this down. — Yeah. Uh — we have options built in to help save on costs. So you might necessarily not want the most highest power model for small things like oh I want to change the color. You don't want this at maximum mode. Um so there's things like uh light and economy and also just we get free sessions with chat GPT. Not just us at Rep with our enterprise license but everyone has a free account. So you can run your prompts, your ideas through uh chat GPT or other um LLM chatbot that you use. In addition, we also have plan mode. Don't forget we have plan mode. It's a little cheaper. That way you can like iterate and think through your thoughts before implementation. So those are some ways to really save on cost. — Yeah. And I think being specific, you know, a lot of people will just like it's not working and so the agent will just run through every possible thing, you know, or like um you know, we're like, hey, — make it beautiful, but what is beautiful? It's so ambiguous. — Yeah. So, I think we talk about this all the time, the context and the spec specificity of the prompts and really like again um especially as the project gets bigger, right? If it's a very simple project and you're like, "Hey, this thing doesn't work and there's only one or two features," it's pretty easy for the AI to figure out like what's not working. But when you have 20 features or five different pages or different parts of the application that kind of do different things and you're just like it doesn't work. It's like well what doesn't work and like how does how do you expect it to work and like what is the desired outcome? And so like again learning how to not just you know like pick the right mode but also provide the right context or enough details in your prompt. And I think that's something that um a lot of people sometimes skimp on or they try to like use real simple prompts where like to get to the next level you probably need to just you know get a little more detailed or granular with those prompts. — Yeah. Start simple. If it doesn't work out then be more specific and don't curse at your agent. — Those are great. — They will charge you. — And going back to Reddit, one of the things I really uh again was a big shift from us uh you know early on in those early days uh again we saw this kind of shift from a previous business model into a new one uh where we are today. And of course the community, you know, it was a tough change for a lot um a lot of us. And um I'd love to hear about how kind of support has changed since then. I mean you've seen it really from the beginning. And I think I always get these like hey no reput support is amazing. I remember our latest uh agent uh for launch like people are like I've had this issue with support. Why am I not hearing back? And people are like no like that shouldn't be an issue. I remember Chris

big shout out to Chris is always like you know has gotten help on his uh support tickets. Like what has changed over the past couple of months uh you know since you joined? Yeah. So, I joined back in October. Um, before me, there was just one person. Now, we're around two. We're hiring more folks as well for our team. So, there's a lot more presence. But overall, I've heard from other folks, there's not a lot of companies that provide support or uh provide or they don't even have presence on the subreddits. And what's also been helpful is that I've been working a lot with producting marketing teams for feedback as well. So not just necessarily customer issues, but let's say you know someone doesn't like a feature or has something to say. Uh the team is really listening. They're looking and learning from everyone's experience. So I thought that was really cool that people at Replet take that feedback and take it to heart. And it's important to note that it's you and someone else on the social side, but we al we have an army of support folks also as well. And that team is growing really fast. They have their own segment over there. And like they're just my best friends. I love them. They're great. So yeah, not only do we have folks like on socials taking your questions, seeing what you're having problems with, but we also have a ton of folks that are seeing those tickets as they're going in particularly for enterprise and pro plans. — Yeah. So you don't have to reach out on social media. Yeah, I mean we already have inapp support if you're a free paying or a free customer or a paying customer. We have like a whole team of folks as you mentioned ready to help. Engineers, accountant, billing folks. They're really awesome. They're really they do really care a lot and um I I've learned so much from them. So, it's not just social, it's a lot of folks behind the scenes as well. — That's a great tip. If you have a problem, please create a ticket. Like I get some complaints and my first question I was like, "What's your ticket number? " drop it in there. Anything else they should be adding like as they're like reporting these issues? Is it just like creating a ticket and being specific about what the problem is? — Yeah, actually, yeah, Kathy, I'd love to, you know, for all of our audience, we try to we, you know, we try to help out as much as we can. What is like the best way to, you know, uh get attention to, you know, an issue that you're having? And also, when you're submitting an issue, what is the team looking for to best be able to help resolve? — Um, so if you're running into an issue, there should be a way in the app to submit a ticket. It's at replet. com/support. Make sure you're logged in with your credentials. If you don't have an account, you can always email us. Um, let's say you don't have access or something. You need account recovery support. com. We're very available. We do use AI to help triage. I mean, we're we have a lot of AI tools. Of course, we need that. And that AI agent is uh very much trained on our entire knowledge base. So, I learn quite a bit from it. So, we interact a lot with it. Um, and sorry, what was the last part of your question? — Well, the details in terms of what they should be providing. So, they create a ticket. It's not just enough like, oh, this is broken. Like, that's a terrible ticket, right? What's a good ticket? — Yeah. Don't say the right. — Exactly. Be specific. — It's ugly. It doesn't work. — Be specific in your request. Just kind of saying um what steps you took, what happened at the time, just helping us repro um reproduce what went wrong. Any pictures will help. And I know a lot of people build a lot of projects. So kind of pointing to that as well because if you're just in there and be like everything is broken um and you have like 20 projects or if you say oh my app is broken, you have like 20 projects, it's kind of hard for us to uh point to which project it might be. So help us help you be specific, provide pictures, give us as much detail as possible. — Yeah. So I imagine it's a lot like you know with the agent, right? as specific as possible with as much context and direction as possible. And I have to imagine it's a pretty difficult task because you're dealing with a very technical thing and Replet allows you to configure like an almost infinite arrangement of potential places where code can break or things can break and so especially with nontechnical users a lot of times like you find yourself in a situation where like something won't work anymore or something is broken and they think it's a replet like platform issue but really it's a code configuration issue and like — how do you determine that or how do you triage those things and like obviously we'd love to be able to support every possible technical configuration inside of replet but that becomes like a an infinite ask but I think that like the customers sometime have this expectation that like um hey it's happening inside of replet and it's not working therefore it's your fault so like how do we deal with that or what's the you know kind of like — yeah I want to say before you reach out to a person or AI agent um again use your if you have access to a chatbot or some sort you can try to troubleshoot there as well. I think having that external perspective helps but first and foremost uh we also have builtin debugging steps um within the workflow depending on where you are sometimes I believe if you try to publish for

example and there are errors publishing there is a button in the app somewhere that says debug with agent or agent even provides you troubleshooting steps in itself it's it will say hey uh can you show me the error log or kind of point me where and I can try to troubleshoot. So, there are some self-service steps you can take and also doing your research. Google it, look it up online, maybe someone else ran into the same issue as you. If you get an error message, they're specific for a reason. So, — take that, type it in, find out what's going on. — And that's a good point is like um you can also ask the agent to log every step of the process, right? So like if you're trying to troubleshoot, you know, and you're like looking for support for a specific issue or you're looking to like help the a to have the agent help you debug, you can ask the agent to like kind of log uh the process, right? And so you can understand where it's failing. And by that, you know, basically um I don't without going too deep on this, but like there's things called console logs. And when something happens inside of your app, the agent or the software will like log specific tasks basically saying this completed successfully or such and such happened or maybe this didn't, right? And so that like something failed and that will log an error or something like that in the console. And so you can ask the agent to kind of log those things for you so that as you're troubleshooting you can figure out where in the process is this falling apart. That will also be very helpful to help you understand like is this a code thing or again is this a replet platform issue and those are two different you know kind of things to triage for where to go for help. — Someone keeps mentioning Quinn in the chat. Who is Quinn? — Who Quinn? — Who's Quinn? — Quinn is the name of our AI support agent. — Yeah. I've never seen anything like it. — So sorry. Can I have some water? — Uh yeah. Um we do we have — I think I'm choking on a strawberry seed. I had some at the open bar. — All right. Quinn is our support agent. Again, it's uh trained on our entire support knowledge base. So, if it comes across like a general troubleshooting or general issue, it should have the steps to help you address it. Uh we do have some weight times, just a warning. Our team is pretty small and lean even though we're growing. So, check out our careers page because we have a lot of uh job openings available. But Quinn is super helpful. Again, I've never seen anything like it at a company before. — Well, we might have to try to get Quinn on the show sometime. Be interested in the chat with Quinn. But in the meantime, people are loving you, Kathy. She is amazing. Or is this Quinn or Kathy? I think it's Oh, yeah. Is it a he or a she or is it like ambiguous? — Um I think we call Thank you so much. Quinn a he — he. Right. So, this is for you Kathy then. She's amazing. Kathy is amazing. — Thank you. Oops. Sorry. — Yeah. I think um I think the folks like meeting the different people from the team and we've had a lot of fun kind of introducing the different uh sides of Replet uh the different people who make the machine run I think and so yeah it's uh it's a lot of fun to uh kind of develop these stories and start introducing the different people to the team and I think that's going to be like a reoccurring trend I think as we start pulling people in more and start doing more of these types of streams. Uh I know a lot of folks are asking like when's the next build or what do you guys you know is this a show like uh you know just experimenting with a lot of different content. It's not always going to be building live with the agent. We're here at uh Saster 2026 in case you're watching and um you know just joining us and wondering like what the heck's going on here. This isn't uh you know the traditional replet live stream and so yeah mixing it up. Yeah, we're um we're on site here at the what is it? San Monteo County Fairgrounds for Saster 2026. uh wrapping up day three of coverage from the event and uh let's see day one we had uh Amjad Mi and Gazi yesterday we uh talked to some folks from the crowd some other folks from Replet uh talk a little bit about how to thrive in the era of AI and today we're talking about what's next for Vibe Code. So I'd be interested Kathy uh to your take like do you have ideas about what's coming next in the space the industry etc kind of as things um evolve here? Um, I know what's coming for me personally in terms of industry. I'm not quite sure. I think it'll become more accessible. Um, I know that Replet's trying to get a lot more non-technical folks like me to try building. I mean I learned a little bit of programming back then and you know some technical skills would be great like learning what a API key does and how databases work is so helpful but honestly Replet really helps me uh overcome that. For me personally, um you know, just based on my own interests, I want to try building some apps for my own gaming stream. And then that could look like uh streaming software or widgets or ways for my chat to interact with me and fun uh customized ways. And also just like making my own websites about the games that I play. — Um you know, there's some websites out there that — aren't available anymore. They're not

hosted. And some of those folks have made it open source. So, I was like, "Oh, maybe I could look into reviving some of those websites. That'd be really cool. " — Yeah. — Making your own games. Maybe. — Well, it's funny. — It's really hard. It's a very hard process. — Well, it's funny cuz the last time we mentioned um you know, I'm playing with some stream graphics also. We should nerd out on that. Playing with some tools that like bring Notion into basically like a OBS um browser source, okay, and turns it into lower thirds and like sidebars and kind of animated graphics that we can use on the show. Um, and the last time we mentioned some of that, like people started sending me ideas of like things they built and how to do graphics. So, it's uh it's cool to see how the community responds to like little prompts like this, things that we think are throwaway lines and then all of a sudden you get like some message from somebody on X or whatever. It's like, "Hey, check this thing out that I built or I'm trying this and um uh I had the same thing with like a Discord bot. I mentioned a Discord bot or a certain thing and somebody's like sending me look what I built the thing that does the thing. " I'm like, and I'm trying to explain like this is really cool, but like we can't let you connect to the server and like to the replet back end to like fix those, you know? So, it's like we'd love to use and implement some of this stuff. But, um, anyway, I'd be excited to see what you come up with on the live stream space is what I'm trying to say in a long way. — Oh, thanks for sharing that. I think one of the best things about Replet is also besides helping people on social media, I also get to see a lot of the cool ideas and apps that people build. So, I'm like constantly inspired. I feel like it's so awesome to see people empowered to build their own businesses, their own tools. So, it's really fun to be on social media. — Yeah. — Very neat. Well, Kathy, thank you so much. — Yeah, you're like our best friend online. So, thank you so much. Appreciate all you do. And you're gonna have to join us again soon. — Yeah. Thank you. — Yeah. All right. Yeah. Thanks, Kathy. Um, okay. So, I think um — we got a couple of we got some p some pin comments here of people making some calls around what's going to happen in the next 12 months. And — yeah, and I think maybe we could just spend the last few minutes uh going with that. And you're welcome to hang out with us. Yeah. You know, — ride or die to the end, y'all. We're here. So, yeah. What are what are people what are people saying? What are people's predictions? What's in the next chat? — Yeah, we had a lot of predictions coming through. Uh yeah, really appreciate um everyone who dropped some of these in. Um, and yeah, one of the predictions, semi-technical operators, let me put this up here. Oops. I think we're all doing it at the same time. There we go. Semi-tactical operators will be most empowered to enable their teams, blending their systems with care and bespoke software, which I think that's kind of what we're talking about earlier, right? Even just on the community team, kind of one of the things that we're doing here. So, Kaiser, um, as always, uh, — I like that term semi-technical because we've been talking about the blurring of the lines between the technical f folks and the non-technical folks and how everybody's a builder. Maybe the new term for is semi-technical. — Well, and there's always this like there's this gap, right? If you're technical and you're non-technical, it's been almost impossible to collaborate, right? Because the technical folks work in these very precise workflows and they're working inside of Git and it's like a it's a black box if you're not, — you know, super technical. And so like be able to jump in there and again maybe not have full access, but to be able to contribute. I remember the first time I pushed code to like a real project that I was working on with developers. That was one of the first posts that went viral on Replet because I was like, "Oh my god, I just contributed to a code base that I never imagined I would contribute to. " And it was a very small update to like this masonry grid. It was kind of like just frontend changes. Um, but for me, it was like this major milestone that I was like, I am now on the playing field with these technical folks in some kind of way. I'm not fully out there. I'm not fully dressed. Maybe I'm just the water boy, but I'm on the field now, you know? And that was like a really cool moment. That's right. That's my semi first water boy. Now you're an agentic engineer, I would say. — I think you're there with Tanner, dude, for sure. — Randy, um, cool. What other hot takes do we have here? — Uh, yeah, I think one from Dario here. Um, uh, agent orchestration platforms. So, systems that manage 20 to 30 parallel a AI agents instances working on complex tasks simultaneously. — We're talking about Replet here. — I mean, yeah, it was cool during the build. We had someone who like somehow was running like 13, right, parallel agents at once. Somebody took down 10k worth of Replet tokens in 24 hours. And people don't know who that was. — I need to know how to do that. Like I'm — Was that for the birthday? — Yeah, for the birthday 10 bill um free day. So like the 24 hours that agent was free like somebody racked up 10k in tokens. But I think um it's possible now replet, but it's not common. So I think Dario to your point this is going to be more and more of a skill that people have is how to delegate a whole bunch of different tasks to these agent orchestration platforms and to be able to do that where you can like get so much done in such short period of time. Right now very few people can do that. — Well and this is what Amad and uh Michael hinted to uh this has been hinted to a number of times um is this kind of this ability for you to you know build more autonomous functionality into what's happening. We talked a little bit earlier about what differentiates like traditional software versus Gentic software and I think again those lines are blurring a little bit. So yeah, the

ability to like kind of build some of that autonomy into or uh determinism or not determinist but um uh what am I looking for the other word um not deterministic the probabistic outcomes in your software — three days of screaming long time earlier today I was like proistic — any thoughts — multitasking is a skill so even though you want to run all these parallel agents maybe some folks aren't too familiar with that So it's kind of like okay what was this task? Yeah I delegated this but I forgot I stepped away from the computer. So luckily we have a combon board but also managing a combon board is a skill in itself. — Uh I think if you're running into these issues uh take give yourself some time um and just you know learn a little bit every day. Take a look you might get frustrated but it is a skill that you need to build just to kind of have all these multiple ideas running. Maybe use a notebook and pen like you got right there. Okay, I got agent A working on this and agent B on this. Um, but it's it can be difficult to multitask. For sure. — That's a good point. As humans, it's difficult to multitask. Now, you got to do it with agents. — Yeah. Another one from MJ. It would be cool if you could have preset agent orchestrations which could then be tweaked depending on the use case. — This kind of sounds a little bit like skills. Some of the skills that developing as well — like having like almost like a marketing agent template or like a sales agent templ or a support agent template. So maybe eventually we take Quinn and he's like a general agent that you can take and you can customize to your own — um company and own knowledge. So something like that I think would be interesting to help kind of people streamline the process. — Something like have you turned it off and on again first of all and then just go through the sequence of steps. — Have you tried resetting your Chrome uh browser? Have you tried clearing the cookies? — Exactly. — Yeah. And the one from Dario, production grade discipline. So built-in testing, code review, and validation processes become standard, preventing the fragile hallucinated code common in coding. — Yes, that's actually probably a really good observation and one of the things we talked about before where it's like, hey, you still have to kind of build some of the guard rails. Your skills file still has to be fairly precise. You kind of have to point it and give it context and the agent uh/models will probably fill some of these gaps and get better at some of that. Uh also building unit tests along the way, kind of, you know, proactively testing the security like we're starting to see in other places. So yeah, I agree 100%. production grade discipline. Uh you know, all of this stuff will kind of be baked in the cake and the whole experience will get more robust and less reliant on your technical knowledge um to kind of be that the catchall, right? Like the agent will be more and more in charge of some of that stuff. — I guess yeah, it's a really interesting topic because like there's like we have a lot of new builders and like getting them to their first prototype like that's great, but then they're realizing there's so much more to like finishing the job. production grade systems. Um, make sure that you're monitoring application while it's deployed, keeping it safe, but also the marketing part, the distribution part. So, it's gonna be curious to see like how much of that like we're going to rely on builders to go up and level themselves up or like joining these communities and watching us do it versus how much of that we're going to bring into the platform itself and help you do it like either easier and maybe automate the whole thing. So like now like recently two weeks ago we launched um was it like our product watch feature where like it monitors your deployed app in production and if it goes down you get an alert and so like the platform is doing that because we're seeing new builders need that but like does that going to extend to marketing and to distribution and other stuff. So it's going to be curious to see what that combination is moving forward. — Yeah. Yeah. Whole suite of um Yeah. — And then another one here from lie uh prediction in two to three years Vive won't just help developers. It'll turn anyone with a vision into a builder. describe the feeling you want and spins up real software that iterates with you. It won't replace great devs, it'll multiply them. And finally, let the rest of us play. This is the future. You're building it right. — We are all builders now, folks. The old titles are going away. Everybody's a builder. You ideulate, you prototype, you ship, you distribute, you got to do everything now. — Yeah. And I want to uh maybe take that one step further, right? Like there's kind of two roles you can take in this world, and we already see this, right? Like you can be a creator or consumer, right? And at some level you need to consume enough to like have good ideas to create. But your role and your desire in this world should be to create new things to build new things. I think this is what like makes human like if you had to kind of at a high level say like what are human beings or what is the goal of human beings? It's like almost always to create something better you know like this is progression. This is evolution. This is like the nature of what we do. It's like why do we build a new iPhone every year? — Mhm. — I don't know cuz we can squeak a little bit more. Yeah. And we should. And why does the model get better every year? Right. like cuz they're just pushing and pressing and like trying to again create something better. And so I think now the barrier to creation or to becoming a creator is being lowered. And I think more and more people should step into that role because you never know the things you create, where they're going to take you, the doors they're going to open, the opportunities you're going to get for kind of getting out onto the field rather than as a consumer, you're just in the you're in the stands, right? You're just kind of watching the world play out in front of you. And I think we're in a moment where like almost

anybody can kind of impact the world around them. And it'd be just better to have more of us on the playing field rather than just watching passively. And so again, we go back to the act action networks, right? Like find an action network. Forget about the attention network. Stop worrying about consuming all this stuff and get out there and go build something cool for yourself, for your community, for your business. Uh you know, like I don't know. — Yeah. — Unlimited screen time for your kids if they're building something. Okay. Like if they're watching something like just like mind-numbing stuff. brain rot stuff. Okay, Nick's at 15 minutes, but if they're on Replet, like give them unlimited time and unlimited tokens. — Well, Marcel, right, a huge shout out to Marcelo on the team. He told us that um his kids are only allowed to play uh video games on the weekend for one hour. So, they can use uh their Switch one I think it's one hour on Saturday and Sunday and then outside of that like they're not all the game. But he gave them unlimited access to Rufflet and you can go do whatever you want. — They built Minecraft. Sneaky, sneaky guy. Which I mean, looks Minecraft. — He's like, I'm so conflicted because they're playing video games, but at the same time, they're now spending more time developing the world. They actually are playing. I'm going to need — like my Yeah, you got to need to connect the Stripe account, kiddos. We're going to like parents are like, you know, there's a lot of these like what do they call them? uh pageant parents where they like — force their kids into like um you know little dresses or little shows and they got to perform on stage and they get like traveling around like maybe now there's going to be like revenue parents or something like they're going to get their kids like you know gaming and like generating revenue for the family. — Good job building that game. Now you got a quota 10 new kids on that game every week or we shut you off. — I'm going to put you on allowance, Dad. like, let me I'm going to generate more revenue as a gamer than I'm going to put my parents on a payroll. — I think what's also great is that Replet can be super creative. It really gets those juices going. Um, I do a lot of art, drawing, and watercolors. So, I feel like when I'm on Replet, which is why I have so many apps, sometimes I'm like, "Oh, I have this idea. " And I just kind of go with it. And I feel like the same way when I draw something. So, um, so don't just oneshot your apps. Like if you make something, iterate on it, refine on it, you're like, "Oh, I think this could look more beautiful. " But you [clears throat] could also be specific. — That's one of my predictions. More creatives on Replet. More creatives using Replet uh rather vibe coding, doing vibe coding. — One last one here from Hain, which is kind of a summary of other ones that we've seen here as well. But definitely see a future uh where Replet evolves from developing apps, developing whole solutions based on agent teams, development, you know, marketing, security, etc. which yeah, I think we've been talking about uh you know, I think it's just kind of one of the natural progressions there. — Yeah. — Neat. All right. — Oh, so many good thoughts and I feel like we could go on forever, but unfortunately I think we got some uh timelines here with the crew in the facility. There's things going on at Saster that we need to go tend to. And we got Brandon in the background giving us one of these. One of these one of these. — One big shout out to our crew. Like yeah, the whole crew, everyone behind us — here helping us like the lighting, the video, the switching between cameras and the and everything's our great crew back here. Thank you so much for making such a great experience. So, hats off to them. — Yeah. And everyone for hanging out with us. We had uh you know, lots of folks come through over the last three days. Um lots of great conversations. Go watch the streams if you missed any of that. We're excited to again kind of just experiment. This was a new thing for us the doing this type of setup. um having this kind of uh you know experience and so uh expect us to do a lot more of that. We're just kind of getting our legs under us as we experiment with live streaming and building these things in real time and honestly being very forward very transparent about the things we're doing which I think a lot of companies don't do as well as us and so we're going to continue to lean into that as much as we can and we think it builds trust as an organization but also as a team. So anyway, thanks everyone for hanging out with us. Any closing thoughts as we go — back to scheduled programming next week. We're going to have deep dive showcases. Um I think we have a partner stream as well. And on Monday we're doing something different where we're doing an exclusive live stream for our enterprise users. Email should be going out. That invite should be going out today. So look out for that and join us on Monday with Cody and Nick. — Kathy, any final thoughts for the crew? — See you online. — Bye everyone. Francisco. Okay. See you. Thanks again for all the die hards. Everybody who was with us all three days. We appreciate everyone. We'll catch you later. Bye. — Awesome, y'all. Heat.

Heat. Heat.