The CEO Behind the Fastest-Growing AI Inference Company | Tuhin Srivastava

In this episode of Gradient Dissent, Lukas Biewald talks with Tuhin Srivastava, CEO and founder of Baseten, one of the fastest-growing companies in the AI inference ecosystem. Tuhin shares the real story behind Baseten’s rise and how the market finally aligned with the infrastructure they’d spent years building. They get into the core challenges of modern inference, including why dedicated deployments matter, how runtime and infrastructure bottlenecks stack up, and what makes serving large models fundamentally different from smaller ones. Tuhin also explains how vLLM, TensorRT-LLM, and SGLang differ in practice, what it takes to tune workloads for new chips like the B200, and why reliability becomes harder as systems scale. The conversation dives into company-building, from killing product lines to avoiding premature scaling while navigating a market that shifts every few weeks. Connect with us here: Tuhin Srivastva: https://www.linkedin.com/in/tuhin-srivastava/ Lukas Biewald: https://www.linkedin.com/in/lbiewald/ Weights & Biases: https://www.linkedin.com/company/wandb/

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  In a world where that we're shifting towards models doing everything and all the value being in the application layer, if there is AGI, the only market that will exist to some extent is what does the model need to do? And the model can only run on inference in just two different parts of inference. So I think on the infrastructure level, I have a workload running across five, 10, 100,000 GPUs. How is this thing going to scale? The second, which is the runtime level problems. Hey, how fast do these models actually run on a given GPU?

• 0:27Lukas Biewald?

  Who's using open source and who's using closed source in your experience? And how do people think about that trade off?

• 0:33Speaker 00

  I think like everyone's on this curve of like maturity. And I think a lot of people start from customer to open source models where they're doing their own things. I think the lower place that other people start is that you go with Anthropic, you go with OpenAI, and you have these great models that can do a lot, but they're either too expensive, have a lot of reliability issues themselves. Or the third piece, which is just like our customers care that we aren't just piping all this data off to someone who's going to train models on this, and that matters to us.

• 0:58Lukas Biewald?

  You're listening to Gradient Dissent, a show about making machine learning work in the real world. And I'm your host, Lukas Biewald.

• 1:07Lukas Biewald?

  Today, I'm talking with Tuhin Srivastava, who is the CEO of Base10 and the founder of Base10. Base10 is currently one of the fastest growing companies in the inference space right now, which itself is a fast growing space. So it's phenomenally successful, just raised a giant venture round. But I was especially interested to talk to Tuhin because he's been at it for a lot longer than people think. I think, you know, you see those graphs where you go from like zero to 100 million ARR in a couple of months, but you kind of don't see the years that I've known him when the growth wasn't there and they're kind of like looking for, for what to do. And that, you know, as someone who has pivoted myself and kind of gone through a lot of struggles, I was excited to talk to him about how he worked through that and how he eventually, you know,

• 1:52Lukas Biewald?

  got the company growing. I was also really excited to talk to him about inference and how it works. I feel like that changes every couple of months. It, you know, everyone calls a space a commodity space, but it certainly doesn't look like a commodity space with the growth you see in certain companies like Base10. So this turned out to be a really interesting interview, went in a lot of directions I didn't expect, and I hope you enjoy it.

• 2:13Lukas Biewald?

  Okay, so I was thinking about this interview. I have like a lot of technical questions that I want to get into that I think will be really interesting, but I kind of thought it might be fun to start with like a little inspiration for people who are like kind of feeling stuck in their companies. Like I feel like you have this like amazing story of like, you know, kind of suddenly taking off after a couple of years of sort of kind of not being the hot company. And I just thought that'd be like a great story to like, you know, kind of share with everyone who might be feeling, you know, at an impasse. Could you, could you, could you tell me about your journey?

• 2:46Speaker 00

  Yeah, it's actually pretty first tech found me. It's actually quite interesting. And Lucas, you could probably, let's write the right word, you could probably like empathize with this is that I've actually been working in early stage companies since 2012, so 2011 actually. And that was like a series of small companies that I was like very early employee at. Most of them were, no, they didn't work. They didn't really work. And then I had another company from 2015 to 2018 that was, you know, I met great friends along the way.

• 3:28Speaker 00

  That, that was mostly, mostly the outside of it, but, and then even like Base10. So Base10 was actually founded in 2019. Honestly, not that long after Wait and Biases was probably founded. And, you know, the world was just very different then in terms of kind of like what the state of machine learning was and what people thought about machine learning. Even honestly, even who we were to some extent targeting, you know, we were, we were trying to work with data scientists, working with small models for internal use cases. I mean, that's what we thought was the market, was the market that was important. And obviously in 2022, everything, everything had changed.

• 4:11Speaker 00

  And, you know, the last three years, we're sure we'll talk about past that in this interview, but. What the reason why I bring up the fact that we've been doing this for 15 years is that I don't think the inputs have changed that much since 2000, since 2000, at least definitely not since 2015. And and really what has changed is probably just, you know, the market, the market kind of carries everything with you. And and the second thing is.

• 4:36Lukas Biewald?

  Wait, sorry, Tim. So when you say you don't think the inputs have changed very much, are you the input? Like, what?

• 4:43Speaker 00

  Every day, like we, I don't think like my motivation towards work, how I think about work, even what I do as much has changed very much, just kind of the velocity of everything around it. Has changed, you know, like, you know, we're still got come in every day and, you know, I check the support channel and, you know, I go and dig into some product problem with some user feedback that we've got. I I still go and try to sell Base10 to a bunch of people.

• 5:12Lukas Biewald?

  Totally.

• 5:13Speaker 00

  Every and I think the desperation honestly from us is even more now than it was when things, you know, weren't working. I think I think the only thing that's changed is that a market arrived and we didn't give up. And, you know, there's plenty of capital to keep us going. And so my my inspiration, I I don't think it should be, it'll be that inspirational, but like my inspirational thing would just be, hey, don't, don't, don't give up, you know, just work with people you like.

• 5:40Speaker 00

  And, you know, follow markets. I think like one of the big things that we did. In the in the early stages of this company was not scale too quickly. And I think like when when markets shift, when markets shift to like a fundamental truths change underneath you. The weight of the company is probably what gets in the way of you being dynamic with those with those things. And like for us, like we were in 2000, from 2019 to 2020.

• 6:13Speaker 00

  Three, end of twenty, when we, when we raised our series B at the end of twenty three, and I think like that's when I was like, oh, we have something real here. I think it was through eighteen people, was through eighteen people. And so like, I, you know, that, that was like a, we were, we were lucky in that, in that way.

• 6:29Lukas Biewald?

  Totally. I mean, do you know my story with CrowdFlower? I mean, I, I also had kind of like a long, like slog before the, the markets go off. So it's like, it takes real skill to kind of keep a company cohesive. I mean, even at eighteen people, I think it takes a real skill to kind of keep, keep people going. And then I think this like incredibly underrated skill, maybe like the core entrepreneurial skill to sort of like jump on, you know, to something, you know, it's an opportunity that's like in, in front of you. Can you, can you maybe talk about like what you were seeing and how you did that? And how much of a pivot was it versus like, you're kind of ready for it?

• 7:01Speaker 00

  Yeah, I, I think a lot of people think it was a pivot. It kind of wasn't. I, I feel like the, we already had a lot of the infrastructure built for it. We were, it's very funny, from two thousand nineteen to two thousand twenty two, you know, obviously no one was calling it inference back then, or we were from the model deployment and model serving. We were going around telling everyone the model serving is commoditized, it's a commodity, it's, it's easy, you know, we'll give that away for free, we'll give that away for free. But we had built all of it. It was, it was just really, what changed was that, you know, if you think about those three things that I said initially, like data scientists, more models, internal use cases.

• 7:41Speaker 00

  You kind of went from like the small models to big models is like one massive market expansion that makes serving actually quite hard. Then we went from internal use cases to production, which meant that like SLA and infrastructure mattered a lot. And then the third thing, we went from data scientists to engineers. And then all of a sudden you have people who can actually have an agency to change the things in front of them. And I'd say it was like around these three truths, it was more of like a refocusing as opposed to a pivot. So we had other parts of the product that we just killed. And we're like, all right, we're going all in on this. But the fundamental product was still there. If that makes sense, yeah.

• 8:22Lukas Biewald?

  Totally. And it's like, what, you know, late twenty twenty three, like what, what actually happened? Was it like one customer that kind of pulled it out of you or what, what's going on?

• 8:30Speaker 00

  Yeah. I do give ourselves credit here. Like, I think we did a good job here. So I think there was a couple of things. So end of 22, so ChatGPT happened, and I think that, you know, that didn't actually have that big an impact on us, except it, one, made people pay attention to AI. Two, and you probably remember this, at the end of 22, a kind of like the narrative just like really shifted around AI. But I think ChatGPT set like two things up, which made it very interesting. One, like it set the consumers, the standards for consumers of what they expect out of AI products. And so like people building things

• 9:11Speaker 00

  who are now our customers, like had like this, you know, this standard to that they had to be doing. It's pretty crazy that the first mainstream AI product was ChatGPT. Cause like, it was so good from like even like user experience. It's like if, it's like if Stripe was the first person to ever do payments,

• 9:31Speaker 00

  that'd be kind of unreal. I think on the other hand, like they also started creating these developer APIs, which again were like high quality APIs. I mean, I think like develop the standards for developers was set very, very high. And so that was one moment that came out of ChatGPT. The, the bigger moment was probably stable diffusion, to be honest, because all of a sudden you had like an open source model that had

• 10:00Speaker 00

  a pro, it still wasn't as good as Dolly two at the time or whatever, whichever Dolly we were on, but it was like approximately as good. And so all, all of a sudden, people started paying a lot of attention, started paying a lot of attention to like the ecosystem that formed around. And I think that's injected a lot of excitement. I'd say, what was really cool is the, the first, I'd say the first use, the first customer that ever came to us. And there's two customers that I can think of actually, who came to us and were like, this is gonna be really cool. The first one was like Patreon was experimenting with Whisper for some, for like generating subtitles. And we're like, oh, that's up. That's very, that's very.

• 10:41Speaker 00

  Very cool. I mean, I think the second piece was, um, we had this, this guy, he was a friend of ours called Seth. Um, and he had fought, he was like in his week, it was his, he took a year off between companies and I think he, he called it his year of yes. Um, and, uh, and, uh, and he, as part of that, he, he wasn't super technical, but he figured out how to fine tune stable diffusion to generate music. Um, and it's called Refusion and he kind of put on the hack and it kind of took off.

• 11:14Lukas Biewald?

  Oh, I remember Refusion totally. Yeah, yeah.

• 11:15Speaker 00

  Yeah. And, and so like at the time, I remember he came to us like the day before his launch. And he was like, I might need a lot of GPUs. And we're like, yeah, sure you will. Um, you probably need five A ten Gs. And I think, I, I, I think at the time he ended up scaling up to like a hundred, a hundred fifty A ten Gs. And we thought that was mind blowing how many GPUs that was, um, at the time. And I think that was like a real spur was like, oh, there's going to be a lot more of this. Like let's, let's really think about that market, um, as a core market. And I think we really, we pivoted, um,

• 11:52Speaker 00

  around, uh, we, we focused around that moment, which was just like, all right, let's change everything. So I think over the course of six weeks, we completely changed the surface area of the product, which is, which is pretty cool.

• 12:02Lukas Biewald?

  Wow. That's awesome. That I mean, good for you for, for making that, uh.

• 12:06Speaker 00

  Yeah. Hopefully you don't have that many customers, but still.

• 12:09Lukas Biewald?

  Still a lot of people, a lot of people don't. Yeah. I mean, I guess like, you know, before I get into the technical stuff, um, you know, you, you kind of immediately jump to like, hey, or something I always think about. It's like, I'm the same person. Like I kind of remember, you know, taking a lot of criticism, you know, when stuff wasn't working and sort of getting praised for like, almost like exactly the same stuff where, you know, you said the inputs are the same. I, I, I just like really relate to that. Um, do you like, is there, you know, any more you want to say about that? Like, do you feel like you've like learned something from kind of watching that experience?

• 12:40Speaker 00

  Yeah. Um.

• 12:42Speaker 00

  I I I think everyone needs to just figure out what drives them to some extent. I think like I I I was call like what I'm driven by is like it's not really about having a successful company as much as it is about like working with people you care about on problems you care about and like everything else will change. And I think it's really easy to focus on the input. When you're like, what do I need to do? It's like, well, I need to go hire more people I like and let me go find hard problems um to solve. Um, I I I'd say um where where I get stuck and like in my and I don't know if you relate to this at all. The darkest moments are almost when you're like trying to company build as the as the core focus.

• 13:28Speaker 00

  Um, if that makes sense, it's like

• 13:30Lukas Biewald?

  No, what do you mean by company build?

• 13:31Speaker 00

  Like when you when you start chasing quarters, when you start chasing quarters and revenue targets, because you know, with this, yeah, with this artificial, um, you know, Gregorian calendar that we've come up with or or like, hey, um, this actually happens a lot with go to market teams, go to market teams um in startups is, you know, when you stop applying playbooks, generic playbooks to things. And it's like, you're not really coming at it from like a like you you you you're almost like trying to um treat everything like the science where I think a lot of it is still the art of it, which is, yeah, hire good people,

• 14:12Speaker 00

  what what work on stuff you care about. Um, yeah, and and really optimize around that. But that can be a bit idealistic at times as well. So.

• 14:20Lukas Biewald?

  I love that attitude. I mean, I I sort of, I think I relate to that when things are small, like, you know, oh, this is just like my project. And I think it's kind of hard to hold on to or it's in my experience, it's hard to hold on to that feeling as a as a company grows. Like, how big are you today? And and are you still feeling like it's a you know, kind of a fun project with people you care about or?

• 14:40Speaker 00

  Yeah, um, look, we're we're a hundred and ten people now. Um, we were, we we have our company offsite next week. Um, and we do it every six months. I think we're we're we're less than half that at our last company offsite. It's growing, it's growing people.

• 14:57Speaker 00

  I think I'm lucky in that I I am a bit like that still. One of the really great things about and about on Base10 is, you know, my my my two co-founders I've known between 15 and 25 years. These are like very close friends of mine. A lot of the people early in the company are still around. You know, even our investors, like a lot of like our Sarah Guo, who is on our board. You know, I've talked to her every day for six years.

• 15:28Lukas Biewald?

  That's amazing, really?

• 15:30Speaker 00

  Yeah, every day. Yeah, every day. And so, you know, like it's a, I think that is, that keeps that stuff alive. I think there's also just a bit of like, I've learned this from some other founder, which is like, I kind of refuse to do things that I don't want to do. And I, I, I think like one of the, like, we have to do a lot of, like, being a founder, I think is actually quite challenging just because you, you, all the problems to some extent fall upon you at the end, but I still kind of like take a lot of agency around like, I would only work on stuff that is business critical for the company, all the stuff I want to do. Like, it's okay if I skip meetings.

• 16:10Speaker 00

  It's okay that, like, that is the, that is the only thing I ask for myself, of myself and for myself in that agency. And that, I think that allows it to feel a bit more like that. Also, like, I'll caveat all this is that everything is really fun when you're growing a lot. And to like last two years, I've felt like that, so, yeah.

• 16:31Lukas Biewald?

  Okay, so we're taking this in a much different direction than I thought. I gotta ask, what's the thing that you don't do that you think someone like me would feel like most guilty about not doing?

• 16:41Speaker 00

  I skip one on one all the time.

• 16:44Lukas Biewald?

  Nice.

• 16:45Speaker 00

  I, I, I, I see one on one all the time and and usually, usually there is of the form of

• 16:55Speaker 00

  Hey, I just, if it's anything pressing, come catch me, come catch me at my desk.

• 17:02Speaker 00

  Um, I, yeah, that's.

• 17:03Lukas Biewald?

  It's funny. I feel like every founder CEO that I've gotten to know well enough, um, actually says the same thing. Like all the way up to like Jensen was actually on this podcast saying that, and I, that kind of gave me like, you know, freedom to feel like I could do that. But it's funny because I feel like it was Ben Horowitz like years ago was like, it's just outrageous to miss like one on ones and everyone just started like feeling like guilty about skipping one on ones. But it's funny, you know, it's like, you know, if it drains energy and like, there's nothing to talk about. Like certainly it doesn't make sense to do like a standard meeting.

• 17:34Speaker 00

  And, and maybe like, I think there's like the, um, the, the things that are like, are rituals which were important. And there's stuff that's just performative at times.

• 17:43Lukas Biewald?

  Totally.

• 17:43Speaker 00

  And like the idea that you need to meet with someone that you, that, that, you know, you work with every day for like 30 minutes to like, you know, reflect or whatever, or problem solve. I think it's like, it's a bit, um, you know, like we, we, we have a own version of that, um, with me and my co-founders, which is that we have these like this weekly meeting. We have this weekly, weekly meeting, um, for like 45 minutes on Friday mornings. Uh, that's, um, where we just, you know, it's, it would seem like a one on one, but it is purely like, we just, you know, we shoot the shit, you know, we, we kind of just do whatever we want, um, in that time.

• 18:27Speaker 00

  Um, and sometimes it works, sometimes it's not. And I think like, you know, that's, that, that part's a lot more important to me from like a ritual perspective. Then like the performative part of like, uh, what are the three things that we need to talk about this week?

• 18:40Lukas Biewald?

  And yet you have a daily one on one with your investors, Sarah. What are you guys talking about there?

• 18:46Speaker 00

  But like I gave it to her so much as like a friend at this point, you know, like, it's like you kind of blown. I, we end up like blurring the boundaries between, um, as, as founders, between everything. Um, and like that, I, I lean a hundred percent into that, which is like, you know, there's no, there's very little boundaries in my life. Or better or worse.

• 19:10Lukas Biewald?

  Interesting. Well, I feel like someone might like, listen to this and think, wow, like base 10 is really lacking some kind of like operational discipline. Do you feel like that's like a fair criticism or?

• 19:19Speaker 00

  No, I, I don't think so at all. I, I think actually like we are, um, It's a, I, it, it, I think we have a lot of operational discipline where it matters. For example, like in our, in our sales force and like the way we do sales, like it's very, very customer centric. And, you know, I, I'd say if you want to talk to our customers, they'd be like, hey, no, no one pays as much attention to us as they tend to. I, I thought they'd say that. I, I, I, I think a lot of the, a lot of the things that we are trying to solve for, uh, like, how do we just do the most impactful thing possible without it draining energy? I think, you know, for, for us, some, at least for engineers all the time, those things are very, very large structure.

• 20:04Speaker 00

  For example, um, which is like, you know, too many meetings, too many one on ones, um, too many, um, too many referrals, too many referrals. I think on the flip side of the, you know, there's like a healthy tension between, you know, what does that need to look like in every company? Which may be different as opposed to just like the industrialization of, of the, of the product org. If that makes sense.

• 20:29Lukas Biewald?

  Totally, totally.

• 20:31Speaker 00

  Yeah.

• 20:32Lukas Biewald?

  All right. So like switching gears to what, what you guys do, I mean, actually I was looking at your website recently, you know, preparing for this and you do a lot more than I knew. I think you're expanding the service area product lately, but, um, I, I sort of think of you as like an inference company. Is that, is that fair?

• 20:47Speaker 00

  That's fair. I, I, I say our north, our north star is inference. That's probably the best.

• 20:53Lukas Biewald?

  It's like, you know, I feel like inference, you know, I, I know a lot of the founders in the space and I've had a lot of friends kind of come in and out of that, um, that category. And I think it's like, sort of like, people feel like it's, you know, like commoditizing. And I think you even sort of said, you sort of thought of it as a commodity, like, I don't know, is inference a commodity service? Like, what are the differentiators?

• 21:15Speaker 00

  Yeah. I, I, I think you need to step back and think about like, what, what, like inference of what? So like, if me, if you think of like generic, generic inference of, of, um, a open source model, like if all we did was serve llama behind an endpoint. I think that's a hundred percent a commodity. Um, and like, you know, it's a, it's, you know, given some quality benchmarks, um, everything, and give us some performance benchmarks, people will go. For the lowest price, I think there's a lot to differentiate on a performance perspective still, but still, in the fullness of time, I believe that, you know, generic inference of vanilla

• 22:00Speaker 00

  models is probably commoditized. I think the truth is for us, like two or three things is that one, dedicated deployment of custom models and fine-tuning models, you know, that is not commoditized. Every workload is slightly different, every model is slightly different, and, you know, serving those models is very, and the tooling around that is quite differentiated. We differentiate on like three different things, I'd say, today, and I think one is infrastructure, and that is, hey, what does it actually take to run this reliably at

• 22:41Speaker 00

  scale? And so we take a lot of pride in, hey, you know, we have many nines, we have many nines, we don't go down, the clouds go down, we don't go down. We've taught everything to be, you know, fault tolerant, pretty elastic, and we align our infrastructure around the customer's needs as opposed to the cloud's needs. So that's the first piece. And I think there's like, we can talk a lot about the infrastructure problems here. I think the second piece is performance. I think a lot of inference providers differentiate around performance. We are very, very performant, you know, in any head-to-head, we don't generally lose that.

• 23:22Lukas Biewald?

  Now, when you say performance, do you mean?

• 23:24Speaker 00

  Tokens per second.

• 23:25Lukas Biewald?

  Tokens per second.

• 23:26Speaker 00

  Latitude.

• 23:27Lukas Biewald?

  Latitude, not the quality of the model.

• 23:29Speaker 00

  Not the quality of the model. Yeah, we mean like, hey, how fast does this move? And I think like, there's a lot of, there's a lot of, there's a lot of talk around how fast these things move. And I think if you, if you normalize over time, what you'll see is that, and if you look at the history of software, that open source runtimes seem to win, all the best quality, all the best performing runtime will eventually just be the open source one, because that's how. Developers want to adopt things and where they want to push them forward. But you still need to meet a minimum of our performance there. We think, you know, we are A plus. We don't think we are. We don't think that's going to win over the in the fullness of time. But you need, you need to be very good investment class there. And the third, and the third one is developer experience and platform.

• 24:14Speaker 00

  I think like we, we think of this as a software problem. We think, you know, most of our customers have between like three and 20 models deployed on base 10. Most of them are serving their customers with different hardware, different scaling needs, with different traffic patterns. And we just provide a lot of software to be able to manage all that. And like, we think that is very differentiated, the same way, the same way that, you know, a lot of other C I C V tools or, or software tools add value. So it's infrastructure performance development. And I think when we look at the intersection of these three, that's where we think, you know, a massive inference company exists in the fullness of time. Now, we do other stuff too.

• 25:00Speaker 00

  I would say those things are very important, but they happen in the, in the service of doing inference. So we want to be the production grade inference company.

• 25:12Speaker 00

  You know, we, we did this, we did this marketing campaign recently, which out of home with this obnoxious buses all over the city, which I thought was great, but it's like inference is everything, inference is day stand. Like that is how we think every day when we come to work, but we, we have like a a training product where we create worthless the training. But again, we, we, we train those things that they can be served on base 10 to some extent, or we, we, we provide fine tuning scripts there. So those models can be fine tuned and eventually be run on base 10.

• 25:42Lukas Biewald?

  Okay, but like. So there's sort of like infrastructure, performance and developer experience. Why do you make this distinction between like an open source model and somebody's custom model or fine tune model? I would think with like, I don't know, like, you know, your newest open source model, all those things could differentiate, like reliability, you know, performance could potentially be better. Development experience, probably you're managing, you know, a bunch of these things anyway.

• 26:12Speaker 00

  Yeah.

• 26:13Lukas Biewald?

  And, and, and you know, even with like a fine tune model or something, it seems like those still could also, you know, undifferentiate, right? Like, I mean, I could imagine like, you know, if everybody's using VLM or something to, to run their models, then maybe it doesn't matter.

• 26:28Speaker 00

  Yeah, I, I think you're right. I, and I think in the, today that is the case. I think you can, you can differentiate the, like, on those things today, and we do. But in the fullness of time, I do think a lot of those things will like, we will know for a given model type, how to run it very fast. And then reliability will have to be, there's like, there won't be, you know, unreliable, the unreliable companies won't exist. And what, and what will be left is reliable endpoints that are pretty fast and where the switching cost between them are pretty, is pretty low.

• 27:04Lukas Biewald?

  Wait, so are you like telling me in the fullness of time, you don't think that you will differentiate? Is that what you're saying?

• 27:09Speaker 00

  No, no, I'm, I'm sorry. Maybe I, maybe, maybe I should separate this as well. I think there's dedicated and there's shared. So what I think about shared endpoints, I do, I think, I don't think there's much differentiation in the shared market. I mean, there's a lot of, there's a lot of differentiation in the dedicated market.

• 27:28Lukas Biewald?

  I see.

• 27:28Speaker 00

  So, yeah, we need, sorry, maybe that wasn't clear, which is like, we are just like, again, if you want to use a Llama endpoint that a bunch of other people can hit, and it's going to do, you know, is, is providing the same tokens per second of the same quality. To me, they are very much the same. And like, that, that the shared endpoint business, 99% of our business is dedicated capacity, which is people have, you know, not multi-ten endpoint, only, only those customers are using those endpoints that they are sending.

• 28:00Lukas Biewald?

  Yeah, sorry. Sorry, no, that's actually even a different axis of like dedicated versus shared, although I guess you wouldn't have a shared endpoint with the, with a custom model.

• 28:09Speaker 00

  Yeah, exactly.

• 28:09Lukas Biewald?

  But. But okay, like a dedicated situation, why is that more possible to differentiate?

• 28:16Speaker 00

  Um, because all those workloads look very different. So like even when you like, look, firstly, it's like, not everything's a language model, but let's assume everything's a language model for now. Which is like, the way that I'm setting up, I'm deploying a model for use case X. Customer, another customer comes along and says, I want to deploy it with this and my inputs are slightly different, my outputs are slightly different. My SLA is different. I can only run capacity in this certain cloud. Um, I need things to be hyper compliant in this way. All those just form different kind of attributes of, of the workload.

• 28:59Speaker 00

  Uh, and that gives you ways to differentiate to serve those customers.

• 29:04Lukas Biewald?

  Interesting. So when I come in and I'm talking to you, you're having a conversation about my requirements and then setting something up custom for me?

• 29:13Speaker 00

  Yeah, or, or we're giving you the tools through our software to be able to configure those things yourself. But, but yeah, so like, you know, you have full, at base hand, when you come out, you have full control of your runtime. We, we can give you the tools to like, make your runtime better, or we can give you some default runtimes, or we can look at the L M, or we can look at T R C L M and tell you, hey, here are some good configurations to run those. But you can do whatever you want there and you might change those configurations based on your runtime. In a, in a way that would make sense for another customer.

• 29:48Lukas Biewald?

  Totally, totally. Okay. So could, could we dive a little deeper into um, how inference works on a modern L M? Like, like first of all, I guess let's talk about how it works and maybe like what the possible optimizations are.

• 30:01Speaker 00

  Totally. So like, I, I, I think like, I think there's like three, the three different parts of inference. This, this two different parts of inference. So like, I, I think that, that you need to think about is, um, you need to think about the infrastructure level problem that you go think about the runtime level problems. Um, I think on the, on the infrastructure level, uh, that is, hey, I have a workload running across.

• 30:26Speaker 00

  Five, 10, 100,000 GPUs, how is this thing going to scale? And so you need to set up your infrastructure to allow inference to be done. So what that might mean is that you need to set up your infrastructure in a way that allows the same user to be able to go back to the same GPU to reuse the KP cache for their workload as much as possible. And so there's an infrastructure, or there might be another problem which we see with some of our customers, which is that I can only use GPUs in a certain region because I want to minimize the number of hawks in the world. So you kind of have this like CrowdFlare-esque type problems or VostelX problems around running inference workloads. And in that way, it's very much an infrastructure problem.

• 31:11Speaker 00

  And so, you know, we've done a lot of work there. You need to acquire capacity. What happens when you need 2,000 B200s? And you've got, one cloud can only give you 500. What do you do? I'm like, how do we kind of give you all those tools to solve that? I think the second stuff, which is a bit more researchy, it's like the runtime level problems, which is like, hey, how fast do these models actually run on a given GPU?

• 31:39Speaker 00

  And, you know, that is, you know, then at some point, like that is where stuff like VLN and TRTLM and SULAN come in. That's why it's where some of the proprietary runtimes come in from some other companies. That's where even like the chip level companies start to come where they're like, actually, at the hardware level, we've changed how inference will be done. And then, you know, the question comes into like, what makes it, what makes this challenging? And so, you know, there's really, like, a couple of things that matter here. I think people care about utilization, they care about speed. And there's somewhat of a trade off there. What does, what do you mean by speed?

• 32:22Speaker 00

  Speed, you know, the general metrics that people care about are like time to first token, which is like, how quick does the first response come back? The second piece that people care about is like the time per output. Token. So after that first token, how how long does each additional token take to come back? And then the second one is throughput, which is a memory question, which is how how much capacity can we put through this? Um, without the degradation of performance. And last one is cost per token, which is like the cost of the underlying um hardware and the KV cache. And how well are you using that? And so there's lots of different things

• 33:01Speaker 00

  um to optimize here. We could talk about those different things. But you know, the first one is like with the time to first token, that's really dominated by the pre-fill, which is the first forward pass through a neural net. You know, time for output is dominated by the decode, which is, you know, the repeated single token steps um that are happening. And then, you know, throughput, a lot of us about memory bandwidth and quantization that comes in, like, how much memory can you use? And how many flops um can you process? And then the cost um per token just comes back to like, again, as I said, like, the cost of the hardware, how big the weights are, how much KV cache are you using? And how much do you have? How much kind of work can you reuse? And there's so much optimization that can go through it. And like, when you come to like some of these

• 33:44Speaker 00

  um open source frameworks that I'm sure you're familiar with and you're just, you're listening to with, which like VLM and SG Lang and TRTLM, um they're all, they're all, they're all kind of um coming up with their own frameworks to optimize across these things. So like, you know, whether that's quantization, or whether that's using flash attention, if you use um kernel, whether that's of continuous patching, the spec deck, like, they they all kind of have the same researching mechanisms to do that. Um. These problems at the runtime level today are really, really challenging and they're, you know, pretty research intensive still. Like, oftentimes, we are pushing stuff that's less than a week old into production, which is pretty, which is pretty terrifying. And probably what's even harder right now

• 34:29Speaker 00

  um that I'm sure you have, you you have heard of, which like the talent that knows how to do this um is pretty limited. And we're just competing with, you know, pretty crazy people for that talent. So.

• 34:40Lukas Biewald?

  I've heard of that. Yeah, I've I've heard that's an issue.

• 34:42Speaker 00

  So so so there's a bunch of different like ways to make LLMs run fast. And that's what I'd call the runtime problems. And then there's the infrastructure problems. We kind of think of both of these things as um interchangeable to some degree, not interchangeable, um both are necessary. Um, I'd say.

• 35:00Lukas Biewald?

  And I guess it's kind of interesting that there's like multiple competing open source projects to do this this runtime thing. Can you, I mean, are there like kind of fundamental differences in opinion that they have that like situations where one works better than the other?

• 35:17Speaker 00

  Yeah, like, but we we love them all.

• 35:22Speaker 00

  We love them all. I I think like a lot of them just come back to like that classic difference of like usability and control. A usability needs to be in control. So like I'd say that, you know, at least from what we have seen at scale, like, you know, who's pretty good at writing inference and Nvidia chips? It's Nvidia. It's it's Nvidia. And so like TRTLM is like the lot like, you know, has a lot is like the lowest level. You know, at scale, at scale, once you, especially with all the new dynamo changes that they're pushing through. At least once it's like elapsed some time from when a model has dropped to when you need it. So like maybe not day one, but like day 90.

• 36:08Speaker 00

  TRTLM is by far the fastest. And you can do them. And that that kind of makes that that makes sense is the lowest level to what Nvidia provides. On the flip, on the flip side,

• 36:21Speaker 00

  BLM, I'd say, and SG lang both, but BLM especially is like, I think they really go towards usability. And so, you know, you can solve the BLM and, you know, honestly, anyone can do inference with BLM. But you're going to take a bit of like a performance hit, you know, you can get around a lot of that by just configuring it. But that's when you really have to become a master. And like I'd say SG lang sits somewhere in between. And kind of, and like, the question might just become. Over time, like, why do all three exist to something said? And I think it's just like a it's somewhat of like nascent C of a market.

• 37:01Speaker 00

  Um, and just slightly different requirements for the people who are using them. And I think like, I help the ecosystem in part like, how? How long did it take for like, front end frameworks to converge? Uh, they're like, we could argue that next day has dominates today. I think definitely dominates today. But what with 30 years into that journey, or like 20 years into that journey, like, it would just take a bit for the market to converge. And I think just given how fast the the ground is shifting underneath us, it's really important that these, all these frameworks continue to exist.

• 37:39Lukas Biewald?

  Do you work with hardware besides Nvidia GPUs?

• 37:44Speaker 00

  Yeah, we, we, we, look, we are the answer is yes. And, you know, we've done experiments and everything that we've served. You know, we've, we've worked with like a lot of the chips coming out of the cloud, who've worked with, we've done work with AMD. We've even done some work with some of the new, like, the new set of chips.

• 38:03Speaker 00

  I'd say like today, my, my, like, what we keep coming back to, like, we love the ball. And again, we want to thrive in ecosystem there is that.

• 38:13Speaker 00

  Cuda's just, you know, like, the reliability and the versatility of, and the developer ecosystem of CUDA, it's just very, very hard. It's very, very hard to like, kind of, um, push, like, to overlook. And so when it comes to like, imagining you are pushing something out into production tomorrow. The last thing you want to be doing is working with anything else besides CUDA. That being said, um, some of like the performance and the cost, the cost trade offs that we have seen with some of the other, um, providers is really, really good and very promising. And I think it would just take a minute for them to catch up here, um, in terms of like the versatility in production.

• 38:56Lukas Biewald?

  And I mean, why is that? Because I guess from where I sit, I mean, I've had in this podcast, like probably five different, you know, companies pitching me on better inference. I like have like probably 15 like other pitches that I've like heard in my life. I go out to dinner and I get like pitched on like, hey, I got like a faster inference. And it seems like a very like testable thing and the market's there. Like I look at that and I'm like, okay, there's no market risk. If you have like faster inference, as far as I can tell, or faster and cheaper. So like, what like, what do you actually like practice rent to? Yeah, yeah.

• 39:32Speaker 00

  Yeah, it's a really good question. I was thinking about that just yesterday, actually. And so like, I think it actually comes down to three or four things. So one is like, look, like it's a, you need the chips. So you need, then you need the manufacturing capability to scale up. And then you need the software layer for versatility. So, and I think once you, and so just like the chips, like you need to be able to know what an inference chip looks like. The second one, the manufacturer capability is like, we need, you know, tens of thousands of them, a hundred thousands of these. How are you going to do that at scale? And three, how do you build like the API so that people can mess around with these things themselves?

• 40:16Speaker 00

  I think all three of those things contribute to this being a very challenging problem for all the folks that you'll probably have on the podcast. But I think the biggest thing, which is, I think there's a challenge and we'll get over it as the market matures. It's just that everything is just moving so fast. It's like, you know, the tip turnaround times are what, like six to nine months? If you're lucky, if you're lucky. And so with that in mind, like,

• 40:49Speaker 00

  The market we are in, like I can't, I can't project more than 60 days forward in this business. Who knows what model will come out or what people want to use. And yeah, and so I, I think that that's speed of the market probably just make things harder. And you've kind of seen the Nvidia right now as well, which is just that Nvidia is like three or four different chips that are like people are talking about right now. So you have the Hopper series, then you have the B, the B 200, the B 300, and you have the GB series. I think customers, like things are moving so fast, like customers are also a bit stuck in terms of like, what, what do we want? I'm sure you've seen a lot of this stuff at CoreWeave, even like when you talk to customers, like people are somewhat

• 41:28Speaker 00

  somewhat, it is moving so fast that the customer is confused about what they want as well. So like, you know, projecting forward nine months is very hard. But I agree. I think in the, in, in the fullness of time, as things, as soon as things

• 41:43Speaker 00

  to slow down just a bit, we'll get there. I hope.

• 41:49Lukas Biewald?

  I feel like in the fullness of time is your, is your catchphrase. I like it.

• 41:52Speaker 00

  Yeah, I like it.

• 41:55Lukas Biewald?

  You know, it's funny, I was talking to a CoreWeave customer recently and they were talking about how they don't like to change chips. And another thing they were telling me is kind of interesting was that they were able to get like over time, a lot, lot more performance out of any particular chips. They, they, I was kind of surprised. They, they seemed like they really did not want to be on the latest chip. So they didn't want to like move their workloads to, to new chips. But I mean, it seems like if somebody's working with you, shouldn't you abstract away the, the hardware that they're running on?

• 42:27Speaker 00

  Totally. I, I, I do think there's like something we, we will move towards over time, which is just like, you know, big, small, extra small, like T-shirt sizing for compute.

• 42:42Speaker 00

  I think that's the case. Again, like, everything has just moved so fast that like, you can't just take a workload that's running on H100 and apply it to a B200 and stick it on a B200 and be like, oh, look, now I have three times the speed up. Unfortunately, like, it just takes a bit of finagling. You know, the flash attention versions need to change to support, you know, FP4 or whatnot, or, you know, the, this needs to have some hardware. And so I think that, that, we will abstract that away. We will abstract that away. I think, again, everything just changes so fast underneath us. It's kind of, it's kind of hard to do that today. And I think what they want changes from a week to week perspective as well.

• 43:25Lukas Biewald?

  Do most customers show up being like, hey, I really want a GB200?

• 43:29Speaker 00

  A lot of people do. A lot of, like, engineers just want, you know, engineers want the latest and greatest hardware all the time. But I think, like, actually, like, we actually see probably two sets of customers. We have customers that show up and like, I have like a,

• 43:47Speaker 00

  give me the, give me the, the best hardware I could possibly buy. Give me the Ferrari. I want the Ferrari. I think the majority of our customers, especially the ones, like, you know, we, we want to sell to companies, like, either in the enterprise or sell into the enterprise. They're a little bit more pragmatic around, like, hey, I need to solve a problem for my customer. And this is like the, this is my workload. This is the size of my model. This is the SLA I need to meet. Could you help me get there? Which I think is probably the better, the, the better way to think about the problem. And it fits back into the thing you were saying about, you know, there is some lock into old hardware, especially if you are meeting the SLAs that you care about.

• 44:23Lukas Biewald?

  Well, I'm really amazed to hear that in your case, because I feel like, you know, in my limited experience with this stuff, there's nothing fun about being on the latest chip with like the least, you know, support and like the hardest to get working. So I, I'm kind of surprised that people want that. They think there's going to be a better performance trade-off, or is it sort of an emotional connection?

• 44:44Speaker 00

  I, I, look, we have, like, I was looking at a customer that we work with, you know, that's doing like hundreds of millions of tokens a minute. Uh, which is just a lot of throughput. Um, I think, I think when it comes to those really, really high volume use cases, those performance trade offs are huge, even, even like with long and video stuff like for video model of B two hundreds provide a pretty, like a forty, fifty percent speed up just off the bat. If you can get it running, um, and, and so like, it's just those trade offs that for their end customers, um, they can drive a fundamentally better, um, user experience, um, with, with better hardware.

• 45:25Lukas Biewald?

  I see.

• 45:26Speaker 00

  Yeah. And, and again, and again, just comes back to a lot of that, um, anthropic, um, OpenAI, Google stuff where it's like, they are setting from like a speed perspective, they're kind of staying the bar for what customer and open to off models need to be able to hit, hit, and they're all, they're all pretty good at that, at that piece of the, the speed part.

• 45:47Lukas Biewald?

  Well, let's take a segue into my next question, which I'm sure people ask you all the time and people ask me, you know, and the, the like, who's using open source and, and who's using closest source in your experience and, and how do people think about that trade off? I'm sure if they're coming to you, they've kind of decided on open source, but I mean, some people ask me, why would anyone ever, you know, not use, uh, anthropic or, or a cheap, so maybe we'll start there.

• 46:11Speaker 00

  Yeah, look, I, I, I think like everyone's on this curve, um, and like of like maturity. And I think a lot of people start from customer to open source models where they're doing their own things. And they want to solve it like, you know, they, they, they're training their own re-ranking models, that they're training their own synthesis models, um, that are, you know, purpose built for what they're doing. Um, and so that's one place to start. I think the lower place other people start is that you go with anthropic, you go with OpenAI, um, and you have these great models that can do a lot, um, but they're either too expensive and they're too expensive, they're, um, the speed's fine, but you know, OpenAI and Anthropic have a lot of reliability issues themselves because again, they're, they're serving

• 46:54Speaker 00

  the massive, massive work for the enormous scale. Um, and and they come to it because they want more control over costs, they want more control over, um, You know, even though on the liability, like give me like dedicated infrastructure at a decent price with some transparency. So I know um what's happening or the third piece, which is just like, hey, we our customers care that we aren't just piping all this data off to someone who's going to train models on this. And that matter that matters to us a lot. And like I I think that's kind of like all three of those reasons are why four reasons, like something custom, something cheaper, something more reliable, their privacy, SLAs um for the enterprise is

• 47:38Speaker 00

  the reason why people shit from closed to open. And I think the reality of the situation is that um everyone's going to use a bit of both. Like I think, yeah.

• 47:46Lukas Biewald?

  Do you have a prediction in the fullness of time?

• 47:53Lukas Biewald?

  Where where this ends up?

• 47:55Speaker 00

  Uh, uh, I don't know, like forty sixty, don't know which way. To forty two, like I I I I think it's all I just gonna like, look, I think there's two worlds that we live in. It's either um there's either um AGI comes, AGI comes, all we have left to do is go on podcasts because we, you know, everything else. I mean, those AGI's would make a better podcast of me than me coming on here, than me coming on here anyway. So like, maybe I don't really get to do that. But, you know, I I think that's a pretty low probability. Um, you know,

• 48:34Speaker 00

  not that, you know, I I I actually think that um models are already very, very good. And the next unlock is probably further away than we think. Um, I I think I think.

• 48:47Lukas Biewald?

  Why do you think that? That's interesting point of view.

• 48:49Speaker 00

  I I I think we're out of data. I I I don't know where the data is. Yeah, I I I don't know where the data is going to come from. Like, I think we've kind of used, you know, we kind of went through this period of the last five years as we kind of ingested more and more data. And so I think there needs to be now an architectural unlock, an architectural unlock. And then then we're really going to like a a research problem again. Um, and I that like, that's my, I think not only that, is that. And that's fine, because I still think we have a thousand X of economic value to unlock what we've already unlocked. So like, I think we could stop today and like still have like more than the industrial revolution of gaining productivity, like which is which is fine.

• 49:30Speaker 00

  But but I think the going back to your original question, the more reasonable outcome is just that, you know, this is going to be a long tail of models and there's going to be and people want that. And, you know, you don't need the the most powerful model in the world to do everything for you all the time.

• 49:50Lukas Biewald?

  Are you seeing reinforcement learning start to change your business?

• 49:57Speaker 00

  Yeah, like I think there's a lot of demand there. I think there's like, again, everything just comes into like a data collection problem now. And like, I think RL in a lot of ways is going to turning into like an inference problem. Even and I tell you, like, we're not, we're not there today from like, we're not there today in terms of like a first class support for those things. But all those things are possible using base ten as a primitive. And so like, I think there's a lot of demand there, especially with folks who are able to connect. And like, this is what's really interesting, working with so many customers at the application layer, is that they're most of them are collecting real preference

• 50:40Speaker 00

  data, like real valuable preference data, which they want to use to make their models better.

• 50:49Lukas Biewald?

  All right. So I sent over my questions. And unlike most guests, you actually engaged with my my pre read and and you added some questions yourself that I was kind of intrigued by, or I don't know if this is your team, but you added a question, why inference is the fastest growing part of the AI market, which seems like incredibly obvious that it would be. But I'm like, wondering if there's something deeper there that you wanted to talk about?

• 51:12Speaker 00

  No, no, I think my.

• 51:16Speaker 00

  I think my team probably out of that, but like

• 51:19Lukas Biewald?

  Good job, team.

• 51:22Speaker 00

  My team really out of it, but like I think actually that's the wrong, probably enough. I don't think, I think it's the wrong question. I think the better, the better question really is like, why is the most important market? Um, like if in a world where that we're shifting towards models doing everything and all the value being in the application layer. Um, you know, there just needs to be a hell of a lot more inference. Um, and you know, like we're, we're very lucky that we can't locked into this, this market or this market showed up, but like it's interesting as a thought experiment, like it might be the final, it might be the lost market, right? Which is like, if, if there is AGI in like, we're just going to need a whole even more, um, inference. And the only market that will exist to some extent is, um, is what does the model need to do? And the model can only run on inference.

• 52:05Speaker 00

  And so I, I think like, that is why it's the biggest market, like why is the fastest running market? I think is pretty obvious, which is that, you know, we're just rushing as a society to like inject models into everything we do. Um, and I think that's like, you know, a, a good, like every day I, I now use, you know, a dozen products that are powered by base ten in some shape or form. And I think like, that's somewhat of an indication of, you know, like how, how big inference is getting.

• 52:37Lukas Biewald?

  Yeah, that's incredibly cool.

• 52:40Lukas Biewald?

  I was thinking as you're saying that tool use also might, I mean, like something, do you, do you also like run the tools for your customers or?

• 52:46Speaker 00

  Yeah. Yeah. And like, and, and I think the other one is like sandbox, like the actual code execution. Like the, the, the, like all these problems are kind of, the, the way I would say it is like, I, I was saying this to someone on our team today, which is like most, um, most, I think most startups have this problem where they, they go into a market and they saturate it and then they start thinking, what else can we build alongside of it? Um, and for like, for me, honestly, it feels like we're just swimming into the ocean and it just keeps getting deeper and deeper and deeper and deeper. Which is really cool, but also somewhat frightening, but like tool use, like sandboxes, um, all, all these things are like RL, RL, like all these things are just inference problems, like, um,

• 53:29Speaker 00

  Once, once they're all built out, and I think that's why I think this market just like continues to get bigger and compound over time.

• 53:37Lukas Biewald?

  I do want to talk about, you know, you, you had this quote about you, you gotta burn the boats, which I thought was, was, uh, was evocative. And yeah, I think a lot of people, I don't know, talk about that, but it's actually like really hard to do in practice. Like, can you, can you talk about burning the boats and how that works and how you get people bought in on that?

• 53:57Speaker 00

  Yeah, we, we, we're just really automational people. No, I'm, I'm, I'm, I'm, I'm, I'm, I'm Jake, yep. Me, I, I, I, I think like, um, I, I'd say the, we're just so future looking, honestly, which is like, we just forward look all the time. Like everything we have built until now is the past. And like what I owe, what I owe the people who work for us, our customers, and even like our, our investors is chasing, you know, what we think is the largest opportunity.

• 54:27Lukas Biewald?

  Okay, great, great, great. But like, what's the most like beautiful boat that you've, you've burned down?

• 54:32Speaker 00

  Oh, we were like, we went into, into, in two thousand, in two thousand and twenty-two, we killed three products out of four in one year. And so like, we'd basically spend three years, like, you know, going back to what you said, like, Basin initially had this like retool type application builder that went alongside the serving engine. It was kind of bizarre that, and, um, that we built all this thing. And we, you know, there was like two dozen odd people who spent two and a half years building this. And I think we just, we just killed it. Like, honestly, like we, we, within six weeks, we were like, let's offboard customers. Let's get them new places to do this work. Um, we're, we're going all on this. Um, another one was like, we

• 55:13Speaker 00

  launched, uh, in two thousand and twenty-two, uh, we also launched the fine tuning product called Blueprint. Um, and it was early for fine tuning back then. Um, and we didn't, like, we, seven people on the company or six people in the company of, of, like, that's a third of the company at the time, spent six months building the thing. Uh, we launched it, it didn't go any of them, it didn't really go anywhere. We realized we had the abstraction wrong in three Three months later, we're like, all right, we're not working on that anymore. And I, and I say, we, we constantly do this as much as possible, which is, hey, rewrite, rewrites a part of the job, throw, throw away stuff is, is okay. And we're actually pretty okay. Even with, let's sprint out this, it's okay if we have to throw it away.

• 55:57Speaker 00

  So like, I, I think like, that's like the, the trade off there is that what it allows you to do is to be a little distracted and go on side quests.

• 56:08Speaker 00

  And, and be like, be okay with that side quest being not super valuable.

• 56:15Speaker 00

  And as long, yeah, as long as you can recenter and go back to the thing you're trying to do.

• 56:20Lukas Biewald?

  Oh, that's kind of interesting because, you know, earlier when we were talking, you talked about sort of like, don't give up and like, keep going. And I, I could see how this two are related, but like, it is a tricky question to know when to, uh,

• 56:32Speaker 00

  I, I said, don't, don't, don't be go, don't, don't give up, but don't be emotionally attached. That's probably the two things I'd say, which is like, just be forward looking to something. And I think like, that's where I've always got stuck when I've been building things. It's like, you know, the worst thing that could happen is you get stuck at a local maxima. A local maxima. And I think my acknowledging, acknowledging that to some extent, I think the second piece, sorry, just to go alongside this is that,

• 57:01Speaker 00

  it, it's so interesting. Like I, I, I know how you think about this because you've been around venture for like, how long? Like 10 years, 15 years around venture funded companies a while.

• 57:09Lukas Biewald?

  Fair amount of time, yeah.

• 57:10Speaker 00

  I, I, I feel like as a venture backed founder, I personally think you're really committing to a type of company you're trying to build. And like, I, I meet founders often who are like, oh, it's, it's not really working. So we're going to start to become cash flow positive. And I'm like, oh, I don't, I didn't think why I'm in the cash flow positive business for a while here. And like, oh, I, I signed up something that's going to be very, very big and just swinging big. I mean, I think like, that is something that I think everyone on our team will, will, you know, we, we try not to take the conservative take on as much as possible.

• 57:44Lukas Biewald?

  It's also interesting that you say that because I think like earlier, like at the very beginning of this conversation, you were kind of talking about how, how did you put it? You were sort of like, I like to think of it more as like this project that I'm sort of doing for this like purpose, like rather than like, you know, kind of like optimizing for growth or something. But, you know, I think like probably your venture investors are thinking, hey, we're working to optimize for growth there, don't you think?

• 58:08Speaker 00

  Yeah, I think so, but I don't think that detached as you think. Because like I think with the, with respect to the former thing, which is like, hey, I need to be happy to win. Like, you know, I'm not thinking about winning, but I need to be happy if I'm going to get there. And they also like, the thing that drives happiness for me is working on big things. And so it's not, it's not necessarily around like, how do I squeeze like the most amount of sales efficiency out of the machine that we have? And, you know, we need a new leader there because, you know, we miss quota of attempt. That's more just like, hey, are we chasing the biggest thing? Are we chasing important problems to some extent?

• 58:50Lukas Biewald?

  That makes sense.

• 58:51Speaker 00

  Yeah.

• 58:52Lukas Biewald?

  All right. I think that's a great place to stop. That was an awesome interview. I really appreciate it.

• 58:56Speaker 00

  Yeah, no worries. Thanks for having me.

• 59:00Lukas Biewald?

  Thanks so much for listening to this episode of Grading Descent. Please stay tuned for future episodes.