How to Build $10M SaaS in 5 Months (2026 Playbook)

Why did Sequoa come and give you $50 million? — They invested before we had AI and we had like maybe 5,000 people using war. — Was it like $50 million when you had 5,000? Yeah. — What? — Today's guest not only left Google but even stepped down as interim CTO at time to start a company that's solving one of the most painful problems in software engineering by fixing the old outdated terminal. In this conversation, Zack Lloyd breaks down how started, how he raised $50 million from SEOA even when he had only 5,000 users. what he thinks about the future of software engineering, the hacks and strategies behind fast growing SAS companies in 2025 and much more. — What happened after AI? When was the inflection point on growth? — Our growth just started going crazy to the point where we were adding like a million of new revenue every week. — Wow. Has that growth broken things at the company? — Crazy problem that I think a lot of AI companies have where we've had a lot of like fraud and abuse costing us, you know, over a million dollars a month. — You spend more on OpenAI or anthropic?

So ladies and gentlemen, um you know, I've been hearing of this super fast growing company called And Zach, you were the CEO of U you know, I remember the time where three or four years ago SAS was this totally different beast, right? There were and we were running a SAS company four or five years ago where, you know, we'd build software, people would come in, spend a little bit of money and SAS, you know, there used to be this thing about, hey, you're going to grow 3x in the first year, 3x in the second year, 2x in year three, and 2x in year four. But today's SAS companies are totally different, right? And I assume the fact that you've raised — uh $50 million from Sequoia and growing like a rocket ship means there's some magic going on. And I wanted to actually tease out that magic, figure out who's your customer, how are you growing so fast, what is the secret? What can I take away? What can the audience learn? Uh all of it in one conversation. So thank you so much for doing this. And you know, I see the OpenAI uh trillion token plaque behind you. So, congratulations on that. Um, you spend more on OpenAI or Anthropic? — We spend more on Anthropic. Um, but it's it's always changing. — And they haven't given you anything. — They actually they actually did give me something. I haven't put it up yet. They gave me I should put it up. They gave me some um they're kind of like book holders. They uh but no, I didn't get like a cool like 10 trillion token plaque or whatever we've spent with them. — I want to tell you how I discovered It's a very interesting story. So, we run a company and you know I was we have a one unit in the company which is the game development unit. We're working on a game and I was walking past one of my developers there and he was using this tool called right? And he was doing like he was just like typing plain English and just giving it a bunch of tasks that you know we would typically run terminal commands for, right? And I was just looking at him and I'm like, "Hey, where did you discover this? Where do you find this? " And he's like, "Somebody told me about this. " And I was like, where did that person find? He's like, I have no idea, but this is useful and I use it every day. And I was like, you know, it's just incredible because if you went back 10 years ago in India, right, in a country like India, we were reasonably slow to adopt software, right? It you would still find Indians using some sort of SAS that was sold to the Indian company by some represent like if it was Adobe, right? There would be some rep in India making sure that the Indian company bought it. But like now we're seeing everything happen bottoms up, right? We're having these, you know, people in companies saying, "Hey, this is the tool of choice. I found it online and, you know, it's AI. It's helping me with my workflows, so I want to get on top of it. " know how long has WAP been around? What did you guys start with? And, you know, where did AI come into the mix? — Yeah, great question. So, we're 5-year-old company. So, we actually predate AI. And the, you know, the original sort of product vision was let's make a uh more modern usable version of the command line terminal. Um, my background as a I've been an engineer for a really long time. I used to be a principal engineer at Google. I've always used the terminal. I've never really enjoyed using the terminal. I think it's kind of like a pain in the ass, frankly. But I've always worked with other engineers who are really good at using it. And so the original concept was like can we just make something that gives that power that um you know engineers who really master the terminal have to normal engineers. Um and so that was the original vision. We actually you know from a business perspective uh the goal was really like or the way that we thought about it was like hey this is a valuable piece of developer real estate. meaning like it's one of there's probably only two tools that every developer uses every day which is the terminal and the code editor and I was very confident that we could build a product that would made the terminal like experience more powerful and more usable so we kind of bet on that um built for a couple years our original like plan for business model was going to be around collaboration meaning like I don't know if are you familiar with Postman do you know that tool — yeah it's from India API develop — yeah from India — so yeah so the original sort of business plan was like let people sort of share uh templated terminal commands or like notebooks uh through the terminal and we actually have that and it's actually it is useful but the business totally changed and we changed the product a ton when um yeah actually we even started putting AI in work before chat GPT came out we had a feature where you could just kind of type commands in English and trans uh translate them to command language but when chat GPT came out we realized Hey, this technology is like super applicable to the terminal where the biggest problem is like it's super powerful but you can never remember how to use it. And so, uh, we kind of went all in on that relatively quickly. And at this point, you know, the way to conceive of warp is kind of like what you said where it's like a general interface for developers where they can just tell their computer what to do and they can tell the computer in terminal commands or they can tell it in English and if you tell it in English, it does it agentically, which is like it's pretty magical. It's like science fiction. It's very cool. — Yeah, I can imagine. You know, when I was young, I used to actually we used to write Ruby on Rails, right, about 10 years ago. And I just remember every time you wanted like to run a server, right? We used to use something called Puma back then, it would be like, "Hey, what was the command? I forgot it. " And then I would go and then I had this cheat sheet that I'd set up for myself. So I'm glad that, you know, life's gotten easier now. Can you tell me and I want to sort of break down the dynamics of SAS here, right, for a second before we get into, you know, before we get

into more, I just want to break down why did SEOA come and give you $50 million? What point did that happen? Was that after the AI wave? Was that before the AI wave? What was the trigger for them? And can you run me through that convers like what's that conversation like? — I'll run you through it because it's probably interesting for other founders. Um they invested before we had AI in the product which is crazy actually in retrospect. Um they invested actually when we had very few users even uh we had like maybe 5,000 people using warp and you know now we're 50 million. — Was it like $50 million when you had 5,000? What? — Yeah. — Like is that — crazy, right? So, — so I I guess like here's how I think of it. So, the sequencing of fundraising for us, which is probably interesting. So, we raised a you know, we raised a seed round or I should just say I raised a seed round before we had any product. I'm a second time founder. Um uh I think I mean I have like a background that I think VCs like right just from someone who was running Google Docs. Uh so I was able to raise capital pretty early and then we as part of that seed round uh had some really cool angel investors come in. Uh one of whom is the founder uh and CEO of Figma Dylan Field. uh and he like loved what we were doing and loved like the development of the product and the early signals we were getting from customers. And so actually he led our series A and then not long after he introduced me to Andrew Reid who's partner at Sequoia who backed us who uh we weren't seeking funding. Uh in fact we weren't seeking funding at the A either. Uh he just liked the concept of a lot and he was I think he did some amount of pattern matching where he's like you know good team building a really hard app. So, one thing is like what we're doing is pretty technically hard to do. Uh, and a potentially very big market and like kind of a weird unorthodox wild bet to make. Uh, and he just was like, I guess he really liked us and wanted to be involved and invested before AI. And so, you know, it was before we had a lot of users and it was way before we had any revenue. Like our revenue growth has really started to like take off in like a crazy way this year. So, and this was like 2023. Uh, so it was a totally it was just like a crazy investment they made, but I really appreciate the bet that they took.

It was cool. — And what happened after AI? When was the inflection point on growth? Because I assume a lot of what we've heard was very recent. — Yeah. So, so I I'll give you just like some stats. This stuff's kind of public. So, going into like May of this year, we were at like a million in AR. So not very high. And then we repositioned uh as like a an environment that was all around agentic development. We call it agentic development environment in June. And our growth just started going crazy to the point where we were adding like a million of new revenue every week. We're now like into the double digits of AR. It's like and it's all happened in four or five months which is pretty cool. It's created a bunch of problems just to be clear. Uh but it's also it's really exciting to have worked on something for you know we spent a long time building the foundations of this app try to make an incredible product experience to go from something that was like growing like okay like we were monetizing it to something that like really exploded in revenue very quickly was a really neat experience and it was like you said I think it's like it's something that can happen in part because of AI and just like the the way that a user you know if you're using Warp before AI, you're like, "This is a much nicer terminal experience. It's cool. Maybe it makes me like 30% more productive. " If you use Warp with AI, it can make you like five times more productive or something. It's like a crazy value ad. So, it's been uh it's been wild.

— Can you tell me what is what are the top two or three use cases that other people are seeing? Like what are the top use cases that developers are seeing that you think really serves? like what are the I'm sure you have power use cases. What are the top ones out there? — Yeah. So, so warp is good for many things. So, we are a great coding agent and there are many people who use it uh like instead of using cloud code or cursor gemini or something like that. So, we're very good at that. We're you know we've consistently been one of the top coding agents on bbench and on terminal bench. And so just like if you want a great coding agent with a very native feeling UI because like again like we aren't a fork of VS code or just a textbased app where like this app is built that really to support the agentic workflow. It's very good for that. But the other thing that's like killer in warp is that you can use it for all of the non-coding use uh like tasks. So if you want to use it to like set up a new project and in the past you'd be like how the hell do I get all these like dependencies installed and it's like it starts breaking and your environment variables are wrong and all that Warp's incredible for that. If you want to deploy it and start doing work that's more like uh in production like with your cloud, I think warp is really differentiated and again because it's like grown out of the terminal is incredible for those kinds of use cases. So definitely far better than our competitors for doing stuff like that. So you know it's great for coding but we can do everything across the whole software life cycle.

— Interesting. How do you how do engineers end up discovering Is it that they round robin through multiple ads or you know not just ads but do you think they look at cursor and then say hey I'm looking for an alternative specifically for terminal like how how's the discovery here? — Yeah I think it depends. So you know because we sort of predated AI we have a lot of people who came to warp for a better terminal experience. So that was the initial draw. In fact, we have a lot of users in India. Actually, India in terms of usage is one of our top markets. I don't actually know how people in India are finding it. Uh like the US and Europe are are biggest ones. I think you know initially it was like uh you know the first day we launched we got like 10,000 signups because we were on hacker news and so just like places where developers hang out they would see warp and try and come and check it out. And then I think after that a big thing for us has been just word of mouth. Like if you want to grow amongst developers I think first and foremost you have to have a product that they really like. Does that make sense to you? Just like product has to be useful. Um and then you know we've gotten a little bit more mature at marketing as we've gone. We only have like a fivep person marketing team, but like you know our YouTube, for instance, has 100,000 subscribers on it, which puts us as one of the top 10 developer companies in the world when it comes to YouTube, which is crazy. We're like, we're not a big company. We're a 50 person company. Uh so I think and it demos really well. So when you see warp, you're like, what is that? Like that is that a terminal? Like but that looks kind of like an IDE. And so you know, all those things have helped us get awareness. Um, but honestly it's still it's a struggle. Like one of the biggest struggles if you're building a startup is just like how do you get people to know about you? I'm sure you know that from starting companies like that's one of the hardest pieces. — Yeah. I think you know the thing about going after developers and doing dev tools is that developers are the internet's early adopter crowd, right? So if you see all the super fast growing even for chat GPT, right? like in the very early days like when they just dropped in like late 2022 um all the early adopters were engineers who said can I use this write code in some way to make myself more productive right so I feel that at least with cursor and you know windsurf and even you guys to a big extent it's this early adopter community that says let me just try new stuff and because you demo so well right and all these tools that demo really well where you're looking at well that's magic right I

think you guys have all seen amazing growth. Has that growth broken things at the company? Like did you expect it? Like how did things change internally when you started seeing growth? — Oh, it's broken a ton of stuff. So the vibe before the growth was like um let's like get things just right and like let's like have a more process or whatever. And then the vibe after the growth has been like, you know, we have like 200 support emails we have to answer. like how do we find someone to help us do that as fast as we can or like you know we had this crazy problem that I think a lot of AI companies have where we've had a lot of like fraud and abuse to the tune of like costing us you know over a million dollars a month of just uh people finding ways to create fake free user accounts to abuse our AI and so it's like we have to fix that that's like a lot that's a lot of money for us and so uh you know all these things that we didn't foresee see as priorities became priorities uh as we started growing. Another big one which like I'm you know I think is interesting to talk about is just like you know this AI based tools do not have the same margin profile as like your traditional SAS tool where it's like a 90% gross margin. like this stuff costs us a ton of money to uh to provide. And so we've had to do a ton of work uh to try to get to a like a pricing model that's actually doesn't lose us money and it's sustainable. And so that's like another thing that's been like it's just like it's a whole different it doesn't matter when you're at a million dollars of scale. It matters when you're at like you know 10 20x or whatever that scale to get the econ unit economics of the thing right. And you know that brings me to a follow-up question with SAS right like you you've hit the nail

on the head that you know SAS now sees a lot more growth than it saw three or four years ago right but at the same time I think the unit economics have fundamentally become very different because SAS was always seen as you build software you have incredible margins you know your cost of creating that extra unit is zero or close to zero so you're able to massproduce software deliver it to everybody else and you walk away into the sunset with these huge margin businesses. — But with AI, because you're fundamentally dependent on tokens, right, from OpenAI or Anthropic or somebody else and you have a cost of this inference, right? They inference that they're passing on to you and even if you end up building your own model tomorrow, you still have that cost of inference. Now, the hope is that over time the cost of inference drops very hard and you know you the margins recover. But do you think that the unit economics for AI SAS companies are sustainable? How do you see that play out? play out? — Yeah, I think right now it's really expensive because you have to be at the frontier of these models to offer the best experience, right? So even right now if you look at any like fixed level of intelligence with these models, the cost per uh cost per token is dropping super fast like it's dropped by like multiple orders of magnitude if you were to look at like a model from a year ago or whatever. The problem is right and that would lead to margin profiles that actually look like traditional SAS. Uh the problem is at the moment you to be competitive you need to be at the frontier and the frontier cost per token hasn't really changed much. Um and it's very expensive. So my thought is that depending on the application over time the bet that we're making is that you probably don't need to be at the frontier. In fact, even for coding right now, it's like, you know, let's say the difference between Sonnet 4 and Sonnet 45, it's legit. Like there's a difference, but like on the evals, it's like a two or 3% difference. And so, uh, if the marginal difference between the Frontier model and like the next two or three models down is not such a big deal in the future, then actually, uh, you're not going to want to be a model company. um if it if they can maintain that big difference uh then they're sort of sitting pretty. So I think the app layer is going to be where it's at. I mean I'm biased just to be very clear like we have an app business but you know it reminds me a little bit of like uh you know there was a time a point in time when like Oracle was like the database you had to use. It's like that's done like that's not true anymore. Like you can use Postgres, you can use MySQL. And so I think if you look at what the um the model providers are doing which is like they are very aggressively moving into the app layer. I think it's out of like concern that like is the model is an is a model API business truly like a defensible thing in the long run. And I don't know my take is I don't think it is but like again I'm biased. — You know I agree with you. In fact, I put out a tweet a long time ago about the fact that the right role to look at the right way to look at many of these model providers is like postresql right or my SQL because it's almost like uh you know when I was very young and I first you know became an engineer I was really worried about should I use this for my database should I go no SQL but then over time I just realized it doesn't really matter it's like a choice preference sort of thing and it's almost like till today I don't know if I've ever paid for my SQL like it's almost when you set up something like a WordPress, it comes free, right? If you want to set up PG, it's sort of you can just do it, right? You can just download it if you need to. Uh I don't think I can't remember paying for it. It's almost like it got abstracted away, right? Even if somebody else was hosting it, I probably paid for it as a packaged sort of deal. Um but I can't remember explicitly and separately paying for it. — So I feel like the same thing's going to happen to the models. Um, but the only

question is if you're if you are now you have to pay that token cost. Are your margins going to be something on top of this because I assume the way your billing is tokens plus a margin. Are you also building a subscription separately? — Yeah. So, we just changed our our billing and I think it's interesting to talk about. So, we what we're doing now and I'll tell you what we were doing. So, what we're doing now is like there's a low price subscription. I think relatively low price like 20 bucks a month. It comes with some amount of AI and then after that you basically pay uh it's consumption based pricing, usage based pricing on top of that. So if you want more AI, you uh you buy more credits. The more credits that you buy, the bigger discount you get and that ensures that like we're not losing money. Uh the and then we also offer something that we will let people do the base subscription and then bring their own API key which is interesting. So if people don't want to buy the tokens through us, we're okay with that. Like I basically want developers to do what works for them. And so as long but they still have to subscribe. So it's still and the margin profile of that is much more like a traditional $20 a month SAS business because just like it's like a good gross margin for us. The way that we were doing it up until a week ago um was we had these like fixed request subscription plans. So, we had like a $20 a month plan, a $50 a month plan, and a $225 a month plan. Uh, and you would get as part of that a certain number of like credits essentially. Um, and that model wasn't working. So, the problem, I mean, it was working from the consumer perspective. They were like this is great because that's how they like want to maybe want to buy it, but uh it's they're used to it. But from us, the problem was to make money on that kind of model, you we're incentivized to have people use it less. So, like the people who we made money from on the $225 a month model didn't use it. And that's like a horrible incentive. Like we want to build something that's super useful for people. And so to get the incentives right, I think for AI companies uh that have a variable cost where tokens is the dominating thing, you probably have to do some sort of consumption thing. Uh or even better if you could do some sort of outcome based thing, but that's not really it's like it's hard to measure outcomes. um in developer productivity right now — and people are always going to lie about outcomes, right? Like it's very hard to track it like there's incentive not to tell people what the outcomes are or when the outcomes actually happen.

— Yeah. — Hey, so a question here a question here, right? Um play this out over the future for me. next three or four years because look, there's a camp that believes that a lot of writing code itself might be automated, right? The VIP coding camp which is press a button, everything gets done. There is the assisted coding camp which is that you know a cursor or you know a and all these tools are going to help me become a 10x or 100x engineer. The goal if you look at the explicitly stated goal of an open AI or an X AI or anthropic their eventual goal is well we want to automate writing code because if you want AGI to come about you need code to be able to you need software write other software right how do you firstly I want to know because you you've done of course Google docs right and it's you've have a long history being a software engineer I want to know your thoughts on how you see this evolve which camp you're in or maybe there's a third camp I don't know about and how is swap going to stay relevant across this because I assume you're going to build more, right? To to enable more of this. — Yeah. So, this is a great question. So, the way that I see it developing is there's like there's three phases to it in my in my opinion. So, there's what I did for most of my career was like write software by hand. Um I think that's ending and the next phase of that is like writing software by prompt. Uh and that's like has two parts to it. So there's like the part that is what looks like warp today where you sit down as a developer and you're like you know literally type or talk to warp and say like build me this feature fix this bug and you sit there and watch it work and you like collaborate with it and review its code and all that and I think that um I think actually we're still at the beginning of the adoption of that despite all the hype that you see like it's going to take a while for developers especially on pro you know like big code bases is to get to a point where they're fully working by prompt. But then the other thing that's happening is automating parts of development. And I fully think that's going to happen too. And um so what I mean by that is like you know uh can you build something where every time a PR gets submitted uh there's an agent that goes and updates your documentation to match the new user experience or can you have something where every time there's a server that crashes you have an agent that goes and investigates and fixes like the crash for you. And so at warp like we're trying to invest in both of those. So, I don't think the interactive part of it is going away anytime soon, but I do think the automated part is really is coming and is a real thing. Um, my guess is like three or four years from now, it's like you as a developer working on a real codebase or spending some portion of your time still like doing this interactive thing, but an increasing portion of your time is like managing or observing or orchestrating what a bunch of these automated agents are doing. Um and that's going to be that's like a crazy future. Uh you know but I think that that's I think that's what's happening. So automation is a real thing for sure. You know I have said

this before right like I' I've gone in public and said that you know this might be the future of software because we all spec we don't know nobody knows what the future is going to be but we speculate on this and looking at everything from you know agent GP GPT many years in the past to everything that's going on with codeex and you guys um it is certain that we will be the orchestrator we'll be the manager of these bots going off writing you know code putting out a PR and then it's our job to sort of orchestrate all of this right but this is a scare scary thing for let's say a 22-year-old engineer to listen to because you know they spent four years at least in India studying computer science uh learning that hey I've got to write every single line of code I've got to make sure I know all the uh you know algorithms like the the top algorithms being used I need to know bubble sort at the back of my head and then suddenly here's this bot that can do a lot of that — straight up and the engineer is like well is my core competence becoming a man I didn't I never signed up to be a manager Right? I enjoy the creative act of solving of writing code of solving problems. And of course, there are actually two different things, writing code and solving problems. But most people conflate the two, especially in India, right? Uh do you think we're stealing away agency from some of these engineers to do this or should they just make peace with the fact that they're going to be a manager of these agents? — It's a super interesting question. Um my take is like I guess there's a few questions in there. So my take is like it's actually very exciting from the standpoint of as an individual the amount of cool that I'm going to be able to build is going to be exponentially higher because of this technology. So if my real interest is in like I want to build deeply useful software for some customer or end user but that's — I'm going to be able that's entrepreneurship, right? That — but that's more entrepreneurship. Okay. So sorry the problem with that Yeah. Go ahead. — Yeah. is entrepreneurship also has this implicit assumption that you're now going to have to go sell and market and build a business and raise money if needed and all of those things. What if I just want to be an engineer for somebody else? — That's interesting. So I I've um you

know I've been like an engineer for — or maybe it's a cultural thing, right? Maybe in India like you know in India we find safety in jobs, right? Like there's a $250 billion IT industry here. uh where we do software services exports to the US and other countries. It sort of holds the fort and a lot of people have their jobs there. The minute you tell people, well, you'll still use the same skill and you have these tools to do things faster, but now you have to take the risk of starting your own thing, that scares a lot of people, right? Everybody wants to be a software engineer and also sell their own software, be like Peter Levels and, you know, Indie Hack and whatnot, but it's also there's so much risk in that it scares people a little bit. Maybe American, you know, the rest of the world has a slightly different culture. I'd like to hear that, you know, that parallel though. — Yeah. So my sense is e let's put aside the founding thing for a sec. I'm obviously like I'm super into building new things, but like I was a software engineer at Google for a long time where my job was um you know build this product Google Sheets, Docs. Uh, even if that were my job, I still would like to have a bunch of tools to let me do it faster and better and and especially take away like the annoying parts of it. Um, and so it to me were like, there's some like I'm trying to get this project set up so I can even start coding or like I don't understand how this new codebase works. Um, and so even if your job is not like you're not out there like net creating some new software and you view yourself as an engineer as more like I'm going to be get tickets that I want to like build and bugs fix. I still think having this new technology is going to be on the balance kind of cooler not worse. Um but the second thing I would say is like if um you know it's not like you're going to not be doing engineering either. So like an analogy that I would use is something like you should still do the four-year CS degree. You should still learn how computers work. code works. Just in the same way it's like as a kid I'm like learning like long division or whatever but then I'm like not doing that. I'm like using a calculator. Uh but there's huge value in uh if you want to like be at all in the world of building software and understand how this stuff works. And like I don't know, I I tend to think that it's not a um it's not a zero sum game, if you know what I mean by that, where I don't see um the fact that you're going to be able to automate a bunch more of this as meaning that people aren't going to have jobs as engineers, if that's like a concern for people. I just think there's like this infinite demand for software in the world. And so even as the tools improve, the amount of software that we'll want to build will improve. But tell like you obviously have a super strong take on this. I'm curious like where your head is at on all these changes. — There is a satisfaction that I used to get as an engineer when I was much younger, right? And I write code I'd solve a problem like I remember you know when I was 20 or something like that uh I was solving the problem of you know I was building like when I was 19 I was building on Ruby on Rails we were building a job platform right so there's so many problems right simple stuff like hey we want to and this was a very different time right so this was a time in India we didn't have stripe we didn't have razor pay we didn't have a payment gateway so you had to work with the bank at the payment gateway like they'd specially sit with you and help you implement they didn't have like a ready to go SDK it's almost like you're working with them to implement a payment gateway, right? So, solving problems back then was like you needed creativity. You needed to find a hack almost everywhere, right? And you need to understand how it worked. — But today, — fun to write the code. Yeah. — Yeah. It's code today. But, you know, we're sitting on a under a layer of abstraction, right? And and I deeply value that because, you know, at the company, we still build software, right? We recently built our own video suite, right? To just manage thousands of videos. It's almost like we built a git for video internally, but it's gotten so much easier, right? And there's so much abstraction for us to sit on top of. And now AI is this new abstraction that makes life even easier. So I feel it has taken out a little bit of the creativity and the fun of being an engineer. But you're right in the sense that it makes you far more productive. We can build more software faster. And you know, I know I can absorb that risk as an entrepreneur and and I enjoy that because now it's more about can we build the thing quick and you know, it's while it's still stable and useful and valuable to the end customer, but can we build it fast because expenses, right? Like the longer it goes, the more expensive it is. So that way I'm super — listen what's do you take more joy in the like writing of the code and getting the code to work or in the like product that the code produces because to me I get the joy of writing the code. It's super fun. It's like solving a puzzle. But the thing that I always like the most and I think there's actually a divide in engineers on this. I like seeing the working product and so I just I find that to be like the coolest thing about engineering is like — No, me too. Me too. — Okay. Yeah. — I love the end product, right? But I I would say that's true today because you know I run a company and I want to make sure that things happen quickly because you were always thinking runway. — But I feel like I when I sometimes when I talk to young engineers, they almost feel like it's a little bit purposeless, right? Like I'm just writing prompts. — Like you'll see this conversation on art Twitter, right? on art Twitter, it's almost like, you know, an artist would come and say, "You're just writing a prompt. You're not doing real art, — right? " So, a person doing, you know, writing a prompt and making art is always feeling a little bit worse about themselves that I'm producing cool stuff, but, you know, I don't have the satisfaction that the art that other artists might have. Uh so this question of craft versus outcome is something where maybe what you're saying is the craft is still there but it's abstracted to more of management and ideas and direction — and but the outcome still matters right you're still judged whether it's humanmade whether it's AI made you're still judged your product is still judged versus everybody else's product your piece of art piece of art um on its own merit so I fully hear you do you think that means that a country like India needs to be slightly more entrepreneurial like or just generally countries that are more interested in just saying hey I write this many lines of code this is my job do you think they should start those sort of people should start being a little more entrepreneurial thinking a lot more about outcome — I mean I'm like I'm all in on outcome and I've always kind of been this way so to me it's like what's the point of writing code right is that you're solving some problem for some user and that has some economic value and like If you are a engineer and you're conceiving of your job as like anything but that I think you're going to have a problem in the future is my opinion. So it's like you you can't you kind of like it's like if you want to be like a car mechanic or something but the whole point is like you got to build something that gets people from A to B. And so it's like I don't um I find that the way the engine works is cool but like what matters is like the actual value it provides. So if I were like running a consultancy in India where the idea has been like we're going to provide code building services I would change that idea extremely quickly to being like we're going to provide um software that pro that solves a business problem. Uh, and like we're going to have to get everyone who's working at that consultancy on that same page as quickly as possible that it's the outcome that matters. Now, I will say this, like I don't think we're at the point with these agents where you can afford to write professional software only caring about the outcome. In fact, I think that's a real antiattern. The agents aren't smart enough. And so if you if that's your approach, if like and you're working on a real software project, you're going to end up with an agent writing code that is like not mergeable or has security holes or um you know causes crashes. And so I I think you have to like take into account that part of getting to the right outcome is doing good engineering still. Uh but I'm always working backwards from like what's the economic value of the thing you're doing? That's just like how the world works for better or worse is like it's the incentives are like to build useful stuff and everything has to flow backwards from that. Agree or disagree? I'm curious like — no I have a term for this right like you can you can say that there's now this new movement of w coding. I actually think there are two kinds of vibe coders. Somebody who is actually a software engineer who knows what he's doing and then can walk in and fix things if there's something that goes wrong. And then there's the other side which I call blind coding. Right? It's the sister, if it's the evil sister of W coding where somebody has no software engineering fundamentals, has never done it in the past — and is saying, look, I'm just going to type in a bunch of prompts. I'm going to get end software. I'm just going to directly ship this to the customer, right? I feel like, you know, it's best to just see AI as a as a tool to make people who are already great software engineers even better, right? The only problem there with the economics in a country like India is if you give the best software engineers these sort of tools, they will take away, you know, 10 roles of the junior folks, right? because that there's that much demand for at Indian prices in India and if you're if you give the top engineers the ability to do 10 projects at once because they're going much much faster uh you know they will swallow that opportunity right which is not a bad thing I think it if you're a good engineer you know this is the greatest time to be alive um I have a question — that is the problem — I I have a question — that's the problem yeah go ahead — yeah why do you feel it's a problem — no I well I feel like if it's If the situation you're describing, which is like you're only going to need senior engineers who are already good, who can wield these tools, get 10x done, there's going to be no need for junior engineers coming in. Then a that sucks for the junior engineers who have spent a bunch of time studying and b it's um there's not going to be a future generation of senior engineers. So it's like you risk losing this knowledge because the way that you get good at engineering is you start as a junior engineer, you work with better people and you learn. So I don't know I like my I I when I think about this particular problem I try not to think about it in terms of senior junior I tend like the people who are still going to have jobs as engineers are going to be ones who have like good like critical thinking problem solving skills who understand computer systems who are adaptable like this technology is changing every month or whatever and so I hope it's not a junior senior divide and it's more of like a you can still come into this And if you are someone who's coming out of school with a great CS education, is really smart and really fast at learning, you can still pick up these tools and be productive. — Yeah, I think it's a high agency versus low agency thing, right? It doesn't matter if you have six months of experience. If you're willing to go use a like, you know, just sit and learn everything at the edge, if you're willing to experiment, build a bunch of things versus I know senior engineers who are just like AI sucks. Nobody uses AI. — Totally. Exactly. like that's a self obsoleting point of view in my opinion. You can't have that uh attitude towards it. And so yeah, it's like whoever's gonna be most adaptable and agency I think is that's the right way of putting it.

— Yeah. I talking about agency. Um can you give me a couple of counterintuitive insights that you figured out running a team and hiring people here versus at Google? — Yeah. So, so we one counterintuitive thing that was very interesting is it just it's connected what we were just talking about is like when we first started getting to a point where warp was usable to build warp meaning like you get this like cool recursive loop of like you're using the tool you're building to build the tool itself. Uh our senior developers were super reluctant. Um and it was uh because they were like the code that this thing produces is not good. Um, I know how to do this better myself. I can do it faster myself. Uh, and so we actually had to do a thing where I was just like, "No, you're going to try to use this uh start every single coding task in Warp within Warp. You don't necessarily have to finish it within Warp. " Um, and then it's totally changed in the past three months where some of the senior developers on our team have like gotten it and are like able to do like wield it in the same way that they could like were awesome at the terminal or awesome at Vim or whatever before. They're now awesome at using these agents. And so, you know, I guess the sort of interesting thing for people who are running companies is like you might have to force it a little. um overcome these biases. You have to overcome this idea that these technologies are fixed in at a point in time because they're not like you might have tried it six months ago and it's it did suck and now you try it and it's you're like what it's it can do these magic things. So I think that's really uh that's really important as someone who's like running an engineering team.

— Yeah. I want you to help me you know sort of tell me where you think this industry is headed. Like I just want a macro insight on AI, right? Because you know how it's going. You know where the how you've been a user of the model since the early days, right? You've seen it evolve. Uh do you think we're going to get AGI in a couple of years where you know press a button and things are done or do you feel like it's going slower than expected? Like give me your updated beliefs, right? Because everyone's had different beliefs about this last year. Now we've seen another year play out. Every year we're getting more information. How do you think the next few years are going to play out with respect to the base models? So, first of all, I don't know, but I'll give you my guess because I think it's more interesting to get a guess than not. Uh, it seems a little bit to me like the latest um the latest iterations of the models have been like slightly smaller incremental benefits than the prior ones. Um, the the story that I kind of see playing out is like I do think the models will continue to improve. Uh, I don't know that if the I don't know if what you'll see is like a exponential improvement or you'll see something that looks more like an S-curve where you sort of on whatever the current architecture is start to see more diminishing returns. I think that's like the gazillion dollar question honestly is like is there some reinforcing loop in the model creation where you get to a point where these things continue at the rate they've been at or do you see something that's more like an S-curve where they start to asmtoically flatten out. If I had to bet just based on what I'm seeing it seems like the latest generation was like not as um big of a step change. Um that said, I still think the other thing I think is like we're very early in the um the like roll out and adoption of like useful AI. Uh it's very useful in coding. um customer service is a second place where there's like a big product market fit. It's like there's this study that came out. That's like 95% of AI pilots at companies don't yield actual benefits. And so I think that the actual adoption and productization of AI for most things is not is just like way behind the what the foundational technology would allow. So even if the models do slow down a bit, I still think that you're going to see a huge boom in AI products uh where like these things are productized in a useful way because that's just we're just starting that. Got it. I have one last

question for you, right? Which is if you were to do the entire journey again, what would you do differently? — Yeah, we uh we spent too much time on our like kind of preconceived notion of what was going to make the business work, which was this collaboration platform. Again, if you're a founder, I think really important lesson is to like listen to the market. Don't get too anchored on your like, you know, your biases going in. I wish we had gone all in on AI sooner, faster, harder. Uh because, you know, we had like an amazing interface for doing something like cloud code, but cloud code beat us to market by like three months with our agentic thing. And like they have all other advantages too, just to be clear, but like we knew we wanted it. we were building it and like had we started executing on it faster we would be further along right now. Um so I would say like don't be too precious about your ideas going in and like I would have adjusted and gone harder on the area where you know there was more market pull faster if I had to do it again. — That's I think very good advice. I mean as a founder you your job is to sort of pivot and identify opportunity and go there. But sometimes, you know, when you're building a startup, you're always like, is this going to distract me going in this direction? Is AI really going to be a thing? And whatnot. But but I think you guys still suck the landing and it's very rare to see and you know, the reason I wanted to chat as well, it's very rare to see a company grow so quickly and be used by people all the way in my gamedev team, which is like all the way across the world. — It's so by the way, it's so cool for me to hear that. That's like — I should have sent you a picture. — Yeah. That's just like that's the thing that gets me most excited about working on this at all is like some person who I've never met in Bangalore on your team is like using the thing that like you know five years ago was just like a Google slides stack where we were like this we think people might want this. Uh and so that that's like the most fun part of actually being a founder is hearing stuff like that. That's super cool. Thank you for sharing that. — Yeah. And thank you so much for building it. I mean, you know, if you've made our team a little more productive, you've saved me money. You've saved us time. — You brought things closer to market. So, thank you so much for doing uh what you do. It matters if if you've ever had that question in your head, which is why are we building this? Are we just building another B2B SAS company, but actually it matters, right? I And I try to tell founders, especially founders who, you know, someone in the team or I'm using their products. I'm always like, don't stop doing what you're doing. — Thank you. Yeah. No. To me, like this is why I think foundings and being an entrepreneur is like the most fun thing and most satisfying thing because like it's a true test of whether or not um your thing only succeeds if it's actually useful. Like there's no cheating around it. There's no like at Google, you can kind of get promoted even if your project fails or whatever. If you're an entrepreneur, uh it has to the only way that you succeed is if the thing is useful. And so I find it like to be such a fun challenge and when it works it's really awesome. — Amazing. Thank you so much for joining us ladies and gentlemen. Go check out and thank you Zach for your time. — Thanks for this was awesome.