Building an AI Startup After MIT: Fireside Chat with Warp CEO Ayush Sharma | MIT Startup Week

Well, maybe we can start with some quick introductions. He's a lot more interesting, so I'll start first. My name is Gabby. I run special projects here at Warp, which means at any point of time, my focus is on the most important strategic initiatives to the company. I was also MIT with Ayush. I was Course 2, which is mechanical engineering, and Computer Science, which is Course 6. And I've worked in tech. I was a software engineer at Facebook, mechie at Apple, Google, and then lots of other places, and joined Warp over the last 6 months, and it's been amazing. Maybe I'll kick it off to you, Ayush. Who are you? And then we'll go into the questions.

Great.

My name is Ayush.

I'm the founder and CEO of Warp. I was class of 2019, computer science here, so of course 6. Also, in any of my free time, I was dabbling a lot in physics, which is what I originally wanted to study. But you come to MIT, you're like, man, CS for the first time. We'll talk more about it, but where I grew up, I had no— like computer science course whatsoever, so coming here it's kind of like changed my life. But it gave me this— it gave me this like—

Make sure it's on.

Yeah, it gave me this very early appreciation for building. So fast forward today, very quick, I'm running Warp and we're 32, 33 people. 33 as of yesterday evening. One person just signed last night actually. We're based in New York, almost all in person. growing quickly. But effectively, the simple version of what we do is we're building this AI-native employee management platform for high-growth companies. So what that means is many of you have probably worked at various places, you've gotten paid. If you zoom out at the company level, turns out there's just like this insane mess of like compliance tasks and manual workflows and like lots of tax-related things and employee onboardings that are just— take very, very complex, time-consuming, error-prone. And what we do at Warp is we are applying really just cutting-edge LLM and agents towards this very old siloed problem space and bringing something really cool to it. So we're happy to talk more about it, but that's the quick version.

Amazing. And so maybe let's start out. Tell us a little bit more about your journey to MIT. Where did you grow up, and how did it get to you? To you today.

Yeah, so I grew up in this relatively small town in India. I say relatively small town because even it's a very high population country, so even smaller places could be really large cities. But I grew up in this town called Kanpur that not many people here would have heard of, but in India it's one of the places that has an IIT. So IIT is like the Indian modeled after MIT's, but they have many of those. So think of it more like the Berkeley system. So they're the top engineering schools in India. One of them, the top 3 in the country, is in my town called Kanpur. So it was a big influence growing up, almost this ever-looming presence into like everybody, all my peers and students and parents, everybody wanted me to go to an IIT, and especially if it could be IIT Bombay.

IIT Bombay.

in Kanpur, my hometown, my mom would have been very happy. But it always kind of gnawed at me because growing up I was just really into physics. I was reading a lot of Richard Feynman lectures on physics. That's like I loved growing up. And so I always had this tension of, okay, I could— and I was actually halfway through high school on a straight track to go to the IITs. There's a grueling multi-year preparation you have to do. For those who are not from the U.S., maybe if you're familiar with a very grueling test-taking thing, this is that taken to maybe its extreme conclusion.

SATs on steroids. It's actually hard.

It's 1.5 million students who take an exam, hardcore physics, chemistry, math, and I think there's about 5,000 out of that 1.5 million. that get into the top schools across the country. So I was probably going to get into, based on all my preparation and scores and everything, to the top 3, which would have been potentially in my hometown. But I kind of rebelled internally halfway through high school. And I didn't tell my parents this for a while, but I started secretly preparing for MIT and coming to the U.S., which was— Interesting choice in hindsight, I didn't know a single person who had gone to the U.S. to study. And so it was very— looking back at it even now, I'm kind of somewhat amazed, it was like, wow, what was I thinking? And so when I did tell my parents that I was actually kind of like behind their back preparing for researching what SATs are, how to write these essays for U.S. schools, they thought I was crazy, I had lost it. they became very kind of like disappointed in me. It's like my child has completely lost it. But eventually they came around when I did get into MIT and Columbia and some news reporters showed up at my home because there's like 3 students in all of India that get into MIT every year, which I also hadn't fully appreciated. I just kind of like went for it. And then so when it happened, a bunch of like national reporters came up and At that point, my parents were probably like, okay, I think he's going to be fine.

Yeah.

So that was probably like one of the more— I feel like MIT is a place that brings a lot of people from all over the world with very unique backgrounds and stories, but I feel like it was a very interesting journey as far as that goes here. So that's how I got here. Let me pause there.

Yeah. So we touched a lot about your journey to MIT. You're here now founding a company. Did you always know you wanted to be a founder or how did you get to that moment?

It's a great question. Let me ask you guys this. Are you all interested in building or joining startups and what's kind of the outlook on that? I'm assuming— I mean, you're here, so kind of some interest, right? Maybe show of hands, are you interested in being a founder, first of all?

Okay.

2, 2.5.

Exactly half.

Interested in joining and working at a startup? More, so almost everybody.

Exactly.

It's kind of a superset of the first question. Makes sense. When I was here growing up and during MIT, I actually never thought that I would start a company. Which I feel like is in a little bit of a contrast to the usual stories that you hear about people who start building in school and they are just like in love with the game of startups, building companies, especially like I think in the past maybe 5, 10 years it's become very accessible in some ways, right? Like I'm sure— I remember YC had come in this exact room many years ago.

Yeah.

I had attended that talk. And ever since they've continued coming in ever more presence. It used to be that that was contained to Stanford because it was just closed, but then I think in the last few years especially as a lot of MIT founders have taken over, they have realized that you can now come to MIT, convince all these kids to drop out or whatever, give them half a million dollars, it's like go start a company. But I never really thought that I would start a company because I first kind of like didn't have any exposure to it. I was very academic— academically oriented growing up, like into physics, math, computer science. And when I graduated, I thought that I would just like do sort of engineering, build things, and very quickly became obvious that I think maybe a couple of months into my first kind of like normal— I had like this—

Internship.

normal kind of like engineering role and I just kind of realized that like working at a big company in— as an individual sort of IC engineer, like it just made me bored very quickly. And it was nice for the— but like initially it was nice because like I first— I was no longer a broke college student, right? Like that was a big thing sort of growing up, like my parents would have wanted me to get a job, financial security and whatnot. So I think in the back of my mind there was a little bit of that, and I— but I realized very quickly that, yeah, that was not going to be for me. And the more I thought about it, it became clear that anytime I was working on something that I truly liked, so building something even in school and outside of school, if I was building something purely out of my own agency and interest, I would stay up super-duper late fixing it, building it, polishing it. Versus if it's just like if I didn't have that internal sort of drive towards it, then I would be kind of lazy honestly. And so—

I can confirm. I was here on campus with him.

And so it kind of in some ways, I mean, we can laugh about it, but in some ways it was an interesting signal for myself that like, oh, when I'm at my best, it usually happens in like in this specific setting, in this specific context where I have full ownership, I have full responsibility, therefore kind of like my ass is on the line, and I have nobody else to point to if something like doesn't work. But also it gives me a lot of like intense energy to just like make things work and fully take the responsibility for it. So it got me thinking and I made the move to join a Series A startup. I moved to New York with Gabby as well. I was there at this startup for about a year growing quickly. This was through like the COVID years. And I got to see a lot of like what happens inside a startup, like as you go from that early kind of like seed Series A phase to Series B and beyond. And once I had some conviction that I wanted to start a company, I just kind of left that role and decided to take the plunge and what would become Warp.

Yeah.

So maybe why, out of all of the things that you could start, why Warp?

It's a really great question. So I, like a lot of college students, I think that one thing that's nice about being in school is that you have a lot of free time to build things. But one of the things you'll notice is that people tend to build very similar things to each other. So they'll— in school there's certainly like a class of problems, there's a class of things that a lot of students try to build.

Give us a few examples.

Okay, so there's this fun concept called tar pit ideas that I think YC has talked about in many of their talks. But what that means is things that are— a tar pit essentially this like landmass thing where from the outside it looks like very attractive, it's like, oh, this is just like fresh water, and you go in and you start sinking. So it's essentially a lot of college students try to build things that are one way or in shape or form kind of like tar pit ideas because they are not exposed to what problems they should be solving. And so while I was in school, I feel like the best thing to do, the best thing that I got out of MIT was just deep building experience, like go build as many things as you can. But I felt like I didn't get a great sense of what problems are worth solving. And so in a typical fashion, when I first wanted to start a company, I actually was like, you know what, I want to build a social media app. And so like every 25-year-old wants to build a social media app, right? And maybe the today version of that is some agentic thing or whatever.

Totally.

But it's very true because people often try to— getting exposed to problems is non-trivial. And for me personally, I tried starting some fun project around social thing. It was really fun for a bit and I was like, okay, But I very quickly realized that I was sort of just doing it because it's like a thing that I knew about, not something that was truly worth solving. So I sort of went back to the drawing board and I started evaluating the problems that I had personally run into in past. So both at the previous startup that I was working at and in the process of actually building this social thing, I realized— I just kept running into this problem of like hiring people and trying to pay them. I realized that turns out the U.S. has— I'll give you kind of like high level— about 1,000+ tax jurisdictions. This is a very specific thing to the United States. Almost no country is structured like this, but U.S., because it's so like— there's a federal government, but the state governments and the local jurisdictions have a lot of mandate, a lot of leeway. So it's a very distributed governance structure you could call it, right? India, for example, is a very different way. Most European countries are also very like centralized in that sense, but the U.S. is very fragmented. So all these tax jurisdictions, as you try to build and scale a high-growth company— hey, welcome.

Hi.

What happens is anytime you You quickly become beholden to all these tax jurisdictions very quickly as a company. So if you're kind of— if you're the founder of a company in the early days, or as the company continues to grow, usually anyone with like a, you know, chief people officer, chief financial officer, CFOs, and accounting people and legal people, they constantly are just like butting heads with these very archaic old .gov portals, accounts, and the only interface there up until recently has been these PDFs and these very annoying websites. No programmatic access whatsoever. So I looked at that. I was like, OK, I think we can solve this finally with how LLMs are getting good at navigating these disparate, disjoint, complex, sort of long workflows. And this was maybe 2, 2 and a half years ago. And actually it was actually pretty terrible in the beginning, right? Back then, if you try imagining long-running agents, the progress there was very abysmal in the start and people were really bearish on it. And so internally, one of the bets we took basically was like, I think this is going to keep getting better. And I think we just kind of have to ride it out and keep building. And both our architecture and our data getting better and our experience, but also just the underlying models getting better. we'll be able to make a lot of progress and in a year, 2 years, it's going to look very different. Fast forward today, that's exactly what's happened. So all that's to say why I started Warp is I had this sort of meandering journey where I tried doing the thing that I think most students or college students or right after school people are aware of, which is do some cool consumer app or whatever, or these days maybe some cool AI app. But I think Personally, it is very underrated if you think of what are some really siloed, disjoint, old things that are not changing in the sense that the need for that is always going to be there and you can bring something truly novel to it now because the tooling has advanced so quickly. So for me, that's what's been really exciting is being able to take something so cutting edge and apply it to— Don't worry, keep going. I don't know if we even need it, but apply it towards this really important business problem that every single company in America at least has to deal with. Right now we're very focused on high-growth companies, so we try to focus on a lot of fast-growing startups and these companies that are a couple hundred employees going into thousands and so on. There I think we see a lot of that, but I think my belief is that every company eventually will want to run on a system like Warp that basically is running itself in the background and you don't have to actively kind of babysit the software.

To that point, maybe the reason why we clicked this is my next question is, we call ourselves an AI-native employee management platform. What does AI native mean in your definition and what does it mean to Warp?

Yeah, it's a really good question. It's one of those things where I feel like there's a short window of time where the word has some meaning to it, but quickly the word AI native will quickly become kind of like for granted, like every company will be AI native. It's almost like Well, you don't say you're a cloud-native company anymore, right? Every company runs on cloud.

So I think—

but in the meantime, I do think that there's some value in differentiating and saying, hey, we are architected in a way that is AI-first. We are taking agents seriously. And internally also, in how we build the company and how we hire people and think about like engineering headcount versus just operations people and things like that, like how we divide the company resources, that is also— there's an AI-native approach to it where you default for using as much AI tooling as possible. So to me, I think that's been the most important thing is just how quickly we've been able to build a lot of the core platform, As compared to something like a few years ago, this would've taken— to give you some context, right, one of the things I mentioned is entire employee lifecycle, and this means anytime a new person is hired in a company, there are so many things that usually have to be done. So kicking off their onboarding and all the way from the initial contracts and agreements to what are all their apps that need to be provisioned, what are the sick leave policy that apply to them, how is that going to work based on like which state and locality they're based in, what happens if they move from like New York to California for whatever reason, like all the underlying sort of policies, benefits, all these things have to change. Get married, have a baby. And so all the way to like an employee leaves and now you have to think about, well, their severance is the first thing and then you have to think about, what happens to their healthcare, there's something called COBRA in the United States which is like you can stay on the coverage for a bit up to a certain period and so on. So in that entire thing, and that's just one employee, right, and then you kind of zoom out and you've got an entire company and your people are always moving, like their circumstances are always changing, and then the regulatory regime is always changing, so some new way of handling tips was passed, I think you guys probably heard the Trump bill changed how like up to a certain amount— Those kinds of things are always changing out there as well. And so managing all of that into a single platform that can automate that is quite a lot of product surface area. And up until maybe a few years ago with AI coming to the scene, this would've taken 40, 50 engineers and tens and tens of millions of dollars to just get the first version of the product out. In contrast, I mean, up until recently Warp was a handful of engineers and we had— before our Series A, which was like 6, 7 months ago, we had raised about $5 million total. And that to me is kind of like the big difference in how we're just building the company. And I think it's something that was just not possible, like you couldn't build an entire platform like this unless you had— 2 years of time, 50 engineers, $50 million in the bank. So like I think there's a lot we could talk about how that— what implications that has for the future of startups and how they're going to be built, but I think, yeah.

Yeah, amazing. I think one thing that's really special about Warp is its foundation in a very contrarian product philosophy. You talked about something called runs itself, Right? One of our taglines is engineered to disappear. Can you touch a little bit more on how you came up with that philosophy, especially when comparing to incumbents and their product philosophies?

Yeah.

So this is a fun one because I think one of the key things that's changing with AI is so much of software was essentially a nice UI layer on underlying manual things that someone had to do. And so for this long arc of history, you can imagine the software efficiency gains were basically coming from the fact that you had digitized some really manual thing, and now because of that, you're able to do it consistently, faster, in a more organized way. And that's basically like if you imagine like hand calculating a bunch of like financial models and then you put that on Excel, well now you've got this like powerful thing and it can be replicated easily, cloned, edited, collaborated, no errors basically, right? That drove the like the first— I call it like the first arc of software up until maybe 2, 3 years ago. And I think that was saturating, like we had sort of reached peak software 1.0 by 2015, 2018. I think there was a lot of fatigue with— that was happening with like classic SaaS software, classic like B2B software, even consumer, like a lot of consumer software had I would argue stagnated maybe even 8 years ago. Like for example, it was the last big social app, right, last big consumer app before ChatGPT. There were only some fleeting viral apps like it was like Locket, there was that thing, right, but these were just like very consumable moment in time, BeReal, very moment in time kind of thing, not really truly sustainable novel thing. But I think 2-3 years ago what started changing And like so many companies have still not realized this, but that era is over, like traditional 1.0 software of you just like digitize the underlying thing and that's kind of it. And I think the next step, the new long S-curve that we're on is this new phase that is super exciting is the software actually just does the work for you as opposed to you are still doing the work but with the help of software. And that shift to me is like such a fundamental breakthrough in how I think to think about products in general. And we're thinking at Warp a lot about what that means for this really big company function called employee management. So we're pretty excited about applying that to all things employee management in general, but I think high level, we think a lot about how can we make less interaction, like how do we reduce the amount of time that the person has to spend in the software, which is kind of like the anti-approach to traditional software where you— because you needed like NPS and you needed engagement and you needed them to bond with your product more, you would make longer and longer flows so that This is very common in— by the way, I was just looking up, came across the other day, apparently in consumer apps—

Yeah.

Like Duolingo and so on, right, they'll have a super long onboarding because it actually is proven across millions and millions of users who have downloaded and they've A/B tested this, that the longer you can make the onboarding without the user just giving up, the higher their probability of retention. And so there's a version of that that even applied to B2B software in some sense where you're not really— like directly optimizing for engagement, but people still wanted this product bonding thing, which is like emotionally you have to get a little bit attached to the software, get used to the software, and therefore incorporate it in your day-to-day because you have to tell people, hey, this is the thing, this is the system that we all use in the company to do things. But I think that's changing. I think what's changing now is The software sort of adapts to you, so it doesn't have to be this very rigid thing. You can ask an AI or an agent, be like, hey, I want to do this thing, how do I do this in this software? So you don't have to just learn the whole UX/UI, and I think that's the direction that Warp is headed and building towards. So it's going to be a long road. I think we've just kind of begun this overall thing, but to me, all software is headed that way. And we think a lot about that at Warp.

Amazing.

Do you think it will get to a point where UX/UI will die?

Well, I'm sure Rafael has some takes on that. My take is that— well, first, so maybe if I can phrase it in a hot take kind of way, it's like there's a lot of Twitter threads about generative UX And I think that's like really stupid because the thing that humans want and as long as humans are using these products, right? If humans are not using these products, actually maybe even then one could argue, but humans want consistency and humans want— they are always like as we are always pattern matching. So why is it that you go to any— website and you have a basic skeleton of how information is laid out. So there's going to be a navbar up here, there's going to be this big title, product screenshot. There's a reason why almost every website does that, because we have been so conventionally conditioned to absorbing information that way, right? So it is Yes, you could go ahead and just make a crazy layout that just pushes boundary. That might be good if you're trying to do a personal portfolio, right? Often a lot of design portfolio pages are really wacky and cool, but you couldn't do that for a website where a customer is supposed to come understand what the product does and then go through sign-up flow or all these things. So backing out, I think, yeah, like a generative UX—

UX—

that always is changing on the user will never be a thing. So I don't agree with that. And so I guess it does come down to the kind of like the thing of will humans be still interacting with software directly? And if they are, I think consistency and UX will still matter, although—

Or maybe it'll matter even more.

And who is interacting too, right? Because probably you will cut out the mechanics, the typers that were doing the input or making some small decisions, and it's now only the leadership.

I don't know. Yeah, like the day-to-day people might get— might not even interact, and it's just like a more high-level thing for like leaders and strategic people. It's possible, yeah. I mean, like if you look at ChatGPT or Claude, what is the UX? The UX is a very simple chat box, but— and I'm sure Rafael could talk about this too— but like there's so much refinement that they have had to do on that chat UX. Like from 2 years ago, if you tried counting all the ways that that simple chatbox UI/UX has evolved to now the little copy button that pops up in the inside container. If you say, hey, draft this email for me, now we'll put, at least with Claude and I'm sure ChatGPT, the email in a separate component and then you can click copy on that specifically, not the whole thing, which— because, right? So there's all these things that they've been refining, and you will notice that as soon as like there's a good UX that any one of these guys come up with, everyone copies it, right? Like every— like, you know, Claude's gonna have it, or ChatGPT is gonna have it in a week. So that consistency doesn't ever go away as long as like humans are trying to use the software, even in its like most basic form. When it comes to like business software, I think I would argue The point of business software is to enforce consistency in processes, so they almost— but once again, like UX/UI will still matter a lot. But I think unless we have a world where things— like the end goal, the asymptotic point where like the thing is truly, truly, truly running itself and you just kind of sit back and give it maybe very high level— it's like, you know, the meme that's like— make me a $100 million company, make no mistakes, right? By the time we go up to that level of abstraction, at that point maybe it doesn't matter, but I don't know. Go ahead.

You're dealing with potentially some of the most important stuff in a company, right? And if you get it wrong, I'm sure that's terrible stress.

Yeah, I've had many sleepless nights, I can tell you about it.

I'm curious, again another design question, how the design shows what's going on? Like if it's all agentic, how do you communicate to the human that's running the company what's going on?

Yeah.

That's a good question. I think one of the things that we have realized is over time there's a level of— there's a right balance of giving the level of detail to the end user, That like if you overdo that, then you're exposing more complexity to them than what they actually care about or are going to find useful. But if you go under that, then I think it kind of undermines the trust, they'll be like, oh, did it actually do that? Like I'll give you an example. Our payroll, like the pay someone flow, it is so fast that Sometimes people, like we legit got people initially who were like, that's it? Like, right?

Brrr.

It's like you could like run your entire company's like millions of dollars of payroll, and it's like maybe 2 clicks, like right? And it's like boom, boom, done. It's like everything's already there, the information, all the earnings and deductions and codes and everything's there. So 1, 2 clicks, done. And we architected, the product in a way that it was actually really fast. And so at some point we internally debated, we never actually did it because our CTO Adam, he was like, no way, that's really dumb, I'm not going to do it. But I legit argued for— in moments in time being like, we should add an artificial 2-second loading thing here with a nice little visual being like, boom, your taxes are being calculated. And like your paychecks are going, right? Because people thought that I was like— if they didn't see that little bit of that, they would think like, oh, nothing really happened. It's just like how humans I guess like perceive.

Yeah.

So to your question, I think there is going to be an optimal point of information of what you should be telling and exposing to the users. And it's almost kind of like progressive— what's the idea called? It's like progressive— basically you just give more complexity the more you want to drill down. I forget the word sometimes, but the concept is basically from video games where the difficulty level, like how complicated of a build you can make in some RPG, in the best ones it starts out very like simple, not with insane deep tech trees and skill trees. But as the game progresses and as you get better at the game, you see more and more and more. So I think that idea is something that is somewhat— it's actually very relevant to building products that do do a lot of work in the background.

Mm-hmm.

So if you do want to drill, if you do want to audit the work of an agent that did for you, you should be able to expand, expand, expand, and be able to audit it. But at the top level, like I think we have to apply some kind of judgment on what is right at the top level.

Maybe one, because I think this will— this fits really naturally—

And we'll come back with more questions too.

Is not only thinking about how do you surface and build trust with an automation, especially in an industry that's very low fault tolerant, right? But I think there's also a question of Because we operate in an industry that's very low fault tolerant, that's very mission critical, the mistakes that we make are very critical to business operations. How do you think about where to automate versus where human touch is necessary?

Yeah. So this is something that I feel like I've evolved my thinking on quite a bit, maybe like the most in the last 1 or 2 years. I now believe that the more you can automate, the better. And this wasn't quite the case when maybe like a year and a half ago when I think it was a bit more open-ended question on how good the models will be and how much you'll be able to go full end-to-end. And what we have seen internally is like the intervention rate on like how often we are finding ourselves having to change something that an AI or agent has like actually already done, it's been on this like very like sharp decline over the last 2 years. And I think it's to a point where I see maybe in like— I clearly see like a year or 2 next year, it might just go to zero. Maybe not entirely zero, it's like 0.01 or whoever, whatever that asymptote ends up being. But I do think now that it is possible that we just are able to completely trust these systems, but not trust in the sense of you are just counting on it without checks, but more that you can build enough validation and testing and evals and Q&A into the system to have that confidence that you don't need that many humans oversighting it. The other thing is I have found it very hard to— it is just fundamentally hard to have great, very smart people just babysitting AI. That's the bottleneck to building a human-in-the-loop kind of system, right? If you think about it, and I think as a result, eventually I think like companies are just kind of being, are going to get forced towards let's just keep improving the underlying system and keep reducing the dependency on any human-in-the-loop component. So my thinking now to summarize is basically I think that the more you can lean on software and LLMs and AI and like code in general, the better it is. And so it's actually kind of like this is a good thing for humans overall. I think a lot of what we're doing is manual, repetitive, tedious work in a lot of these roles. And I think one example is if you look at any of our large incumbent competitors like the ADPs or Paylocity, one fun fact about them, if you look at kind of like their headcount, like 30%, 40% sometimes of their headcount is in functions like operations and tax and finance. And there's just essentially humans, like lots and lots of them. And the company is basically this like giant ops labor thing where like lots of them are in like these low-cost countries and they're like double, triple checking, doing a lot of tedious work. I think that we are going to be able to like do more interesting things with our time when we automate the more tedious. It's not all positive. I think there's some like still I know there's like underlying question of like what will like humans do like as we continue to go up the stack of like, okay, well, this was more repetitive and now this is going away and now this is going away. That is a more existential threat, but I don't know if that's— we chat about this separately.

Yeah.

Yeah, I'd be curious to know, you mentioned on the slide as well, right now your focus is on high-growth companies in general. I'm sure as headcount is growing, also the company is going to evolve, and what the difference is working with a high-growth company versus more of an incumbent, and how you think about going forward what that is going to look like. So if I understand the question correctly, it's something like, if we're focused on high-growth companies, like they evolve and grow quite quickly and so the product kind of has to keep up with that, right? And how that differs with just like normal general like companies, like take like a manufacturing plant or something. So what's— that's a very interesting question. I mean, the reason in startups like you always have to kind of have a bit of a directionality to which customers or which sets of customers are going to find the most value early on. And so to that end, and this is something that's like not super obvious to— like if you haven't tried starting a company yourself, and one of the first things you try to do is you try to really narrow down which sets of customers will quickly find value because guess what, your when you're just starting out, it's like maybe you are a couple of people in a WeWork or something and you've got this like first version up and running and you somehow— in fact, in our case, we're asking you to trust us with like all your money and like your employee data and all these things. So we had to kind of pick our focus and we had— I think in hindsight we made a really good choice betting and having a conviction on Oh, we're going to go solve for this startups and high-growth type companies because from first principles they just experience a lot of that pain in a compressed amount of time and also don't have good staffing or humans that are dedicated to finance, HR, accounting that a large enterprise-scale company is going to have, so at least they've got it covered. with humans just working on these things countless, 24/7. But it really— the other side of that to your question is like because these companies grow so quickly, we are just like we have no choice but to grow with them or we'll lose them as a customer. And like I've had maybe one customer out of now lots and lots of customers that we lost like that and I was like this is the worst thing, we cannot have this again, how do we just make sure it never happens? So we're constantly taking this scalability point in the platform very seriously, just like how do we— like if a company is going to scale from 10 people to 100 people in 9 months, which we have seen happen with some of our customers, so that's like on an absolute scale 100 people is like okay, it's manageable. But it's actually the growth of going from 10 to 100 quickly while you don't have all these processes as a company, you don't quite know what's best practice versus not, and Warp still has to just work across that entire curve. That has taught us a lot, especially on how to forge and kind of carve out a really strong product. There's a lot of fluff in our incumbent vendors that just is like stuff that you don't really need, but they figured out ways to sell it, so they try to sell it to you. But we've had to really just like hone it down, narrow it down to be like, what do companies actually need and how can we do them like exceptionally well? And I think as we are now starting to go towards more broader companies, now we actually serve we do serve normal, like all the way from like small-medium businesses to now we call this like growth segment, like a few hundred going into 1,000 employees and so on and above. We are finding that those customers have similar underlying kinds of like, oh, too much like compliance work and manual like state taxes and all these things, and our product is the best one to solve that because we've been forged kind of in fire in some sense. So it's— so far we're finding that as a unique advantage to how we've built the company.

Any other questions or we can go back? Okay, great. It is 7:50.

Have time for maybe one or two questions and then we're going to—

One or two questions or— yeah.

Go ahead. If you could just do MIT again, all those years between MIT and where you are now, is there anything that you wish you had learned or is there anything you would have done differently? Yeah, this is always a great question. There are always— looking back, you can always figure out things that you should have done differently and I will give you one or two of those. But I also— what year are you, by the way, right now? Currently a senior. Okay, awesome. So you're kind of coming towards that. So this applies to you too, right? But I think looking back you can always be like, oh, maybe do this differently, et cetera. I do find that there's a very famous inspiring speech, I think Steve Jobs, right, like looking back, connecting the dots, and he talks about how he took this calligraphy class that at the time seemed seemingly random and then he is able to kind of tie it all the way to when they were choosing the fonts for the first Mac and it was like iconic and like he was able to draw from that experience. I do think that I find that really— like I resonate with that a lot because I studied computer science, I built a bunch of things, I studied physics in my free time, I studied a lot of like human psychology, cognitive science, neuroscience that gave me an appreciation for how we as humans think and feel and what makes us tick, maybe at a more mechanical level, evolutionary, why are we wired this way we are. And all these things in hindsight I feel like I'm able to draw on those things now that I'm building a company, building a startup, And I can apply that towards how I think about products, I can apply that towards how I think about building and scaling a company and an org. And so I find that that's a great view to have. I think every experience you're having, it's going to shape you in some interesting way that over time you'll find out and be able to draw from. That being said, I think maybe Maybe one thing that I feel like more students should do in school is just— maybe it's somewhat like everybody says this, but it's very true— just make more use of the fact that you're in this sandbox environment with no real— like true responsibilities. And in the moment, it doesn't feel like that at all. That's why everyone says this, but no one actually like is able to incorporate. Because in the moment, you're like, oh my God, all these classes. I've taken like 5 really tough courses.

And finals.

And I've got all these pieces and midterms. And it feels like you've got the weight of the world. And now you have to figure out internships and jobs and all these like, what do I want to do, right? All these things. But like, man, It's a construct of our imagination. It's like we have constructed ourselves this imaginary— not like somewhat real, but—

Mm-hmm.

It's largely like this weight we have taken on ourselves. But in hindsight, you will always be like, man, this was a sandbox environment. I could have learned anything, took any class, build anything, use my free time however I wanted. I think taking more advantage of that, of just pursue whatever interest you have, go deep on any rabbit holes you can find, and just be like treat this as these are your years to just do that and you'll always have the rest of your life for more professional career-oriented things and whatnot. So that's my one maybe piece.

Yeah.

Anything, final question?

Yeah. Okay.

Last 2.

One or 2, yeah.

Mine is about— you said that you found your company idea like later on after MIT. Why do you think so many people like nowadays are— at least it looks like it, I'm not sure if it's true— but like loads of kids are just dropping out of MIT and stuff?

So this is where it's— the simple answer is because I think startups have become a little too glamorized and there is a lot of— glorification of startups and founding. I think it used to be that starting a company and being a founder and entrepreneur was kind of low status, but unfortunately— maybe it's fine overall, look, maybe it's a net positive that more people are starting companies because we get to have more people starting new things, right? That's great. But I do think some amount of fact that it became a little too glorified I think on Twitter you will see people kind of talking about like 996, this, starting a company, dropping out, gone into YC. I think this is, in my opinion, it is the wrong way to orient yourself towards building a company because here's the reality, I can tell you right now, it is absolutely terrorizing to build a company. and found a company. It is day in, day out, you're going to have sleepless nights, you're going to have like heart pounding on days, you're going to like think like you're really, really dumb, there will be days that just suck. And I wish that people talked more about it. It is very, very hard and the highs are very high but the lows feel like man, this really sucks, right? And personally I feel like the whole thing has just like made me Yeah. More of a stable, just like I can kind of now weather any storm kind of a person because I've just seen so much in the 2, 2.5 years now. But coming back to your question, I think the answer is because it became a little too glorified, a lot more people started doing it. And I think that that is in general, I think back to my previous answer, in school just like go deep on anything you want, study anything you want. That is probably a better use of time than trying to start a company. A lot of people who start a company while in school, they are neither starting the company 100% nor studying 100% or like being present in school 100%. So you're just kind of— I think you end up diluting that. And then if you drop out, then it's like, okay, well, that was your one time in life to experience this like sandbox-free— get connected, talk to all these like really smart peers. You have this incredible place full of brilliant people, like meet with them, learn from them, et cetera. And for years to come you'll be able to like work with them in various capacities. So these are the things that I don't think you can get outside of a place like— especially if you're lucky enough to be somewhere like MIT or— around here. So I'm actually— all that's to say, I think that we should talk more about how hard it is to start a company. People really glamorize it. And actually, you know why? One of the things you'll see now, so many people start companies, but 7 months later they've moved on. They're like, okay, I'm shutting this company— like so many like eulogies of startups like Oh, how long were you at it? 6 months. Like entire Gen Z companies that get formed and dismantled before like the calendar year is over, before the news cycle on Twitter is over. So I think it kind of is like a glimpse to like, no, if you're going to be doing this, this is a long-term thing. This is a 7, 8, 10-year thing.

There are—

quick successes, but they're very, very, very rare. They're very outlier. And because they're on the news, you think that's how it is. No. How it is is that most companies die. Most companies just die. They don't get anywhere. And then the ones who do, it's a long, hard journey. So I wish we talked more about it. And then when it comes to school, if you're in school, just like make the most out of being in school. You can always start a company. I think the best thing you can do— Paul Graham has a great talk I recommend watching. I think it's like how to prepare yourself to start a startup. And he's basically talking to college audiences, so I recommend that talk a lot. And I think the TL;DR there is basically immerse yourself in as many domains and as many problems as you can and go deep in things because that is what is going to be valuable when you eventually— if and when you actually want to start a company. So I'll end there.

Amazing. Final question?

If we have time. You mentioned the importance of finding the problem you want to solve and the niche market that you need to go after. And this at the beginning is critical and I agree. I have been grinding with that. In the journey— so from what I've heard, it seems that you hit the jackpot Both times with those big decisions, have you had more points in time where you were evaluating a pivot of focus of—

yeah, any changes of direction? So for us, as I was mentioning, initially when I decided that I was going to start a company, I started working on a social app. And so in a sense, the first thing that I went for, it was like completely wrong and while I had a lot of fun doing it and it was good to like build something and get it in the hands of actual users who really liked the product for whatever reason and had a lot of fun on it, I had to completely scrap that thing, be like, no, this is not the direction of where we're going to go. But I think what you're trying to maybe get to is once we committed— so that was maybe like the first 2, 3 months and we kind of did it as just like let's try it and then quickly pivoted away. One thing, like in general when it comes to pivots and I think the best pivots— I think Elad Gil, who's one of the early investors in Warhol, so he talked about it as the best pivots he has seen are where the idea, like you change the market entirely. Like you go from whatever you're doing to completely different. And the most common failure for when people try to pivot is like they'll slightly change the— The product. Like they'll slightly change the problem statement or the user case or whatever and they try to do it. And the reason it doesn't work is because like the general market they're in has something about it that makes it a bad area for them to go after. So in our case, I think we were in the sense that like fortunate to pick a market that is really, really big and continues to grow and is not going away. And also happened to be kind of like at the edge of this new technology coming up where we took the bet of like we'll be able to use it to build a much better system here. But on a more— I think you zoom out and it kind of looks like this, right, things are going up and kind of steadily growing, but on a more micro it's always like that. So even right now, We're 30+ people quickly growing. We went from 3x the headcount, the team size in the last 4 or 5 months. We're rapidly scaling. I think we found this great office in New York a couple of months ago. We're probably going to outgrow it soon. But at a day-to-day, I'm still like, man, this seems tough and there's all these problems. So it is a bit of a constant thing. I think that's a good thing. My sense is listening from— like I was listening to this great talk from Frank Slootman who is now the chairman but was up until maybe last year the CEO of Snowflake. He's very famous for just like being this very hardcore intense— he's Dutch, very kind of this like determined guy, right? And he talks about it as like there's no end to it, like you can go to the Snowflake level $100 billion IPO, that guy you ask him what's going on, he's going to say everything is fucked. Right? Like it's screwed. Like there's all these things that are going horribly and like these are all the things that we could die from and it's just a disaster, terror, terrifying things day in, day out. So I don't know if— he's not exaggerating in this sense, but like yeah, it is just like the reality. Like if you're going to be doing this, there is always going to be like some amount of doubt, questioning, because we don't have perfect information. If we did, we would know and then like the investors would just invest and like always like there will be just like 10 companies a year, but that's not the case. We don't have perfect information and a lot of it comes down to like how much determined you are to just like keep going and with some amount of confluence of luck. So I think that's— we're out of time a little bit, but thank you everybody.

Thank you guys so much.

I appreciate you all being here.