S4 | E19 | The future of software engineering teams with Scott Breitenother - Co-Founder @ Kilo

Speaker A: Foreign.

Speaker B: Welcome to the Think Data podcast, brought to you in partnership with dataworks. If you want to stay up to date with the latest breakthroughs and trends in the world of data and artificial intelligence, and if you're curious about some of the strategies that companies and founders use to launch data and AI products, then you're in the right place. Our aim is to bring together a diverse lineup of fantastic guests, from the founders through to accomplished leaders and product owners, uh, at some of the most fascinating data and AI companies worldwide. They will each offer you their own unique insight into what it takes to launch and scale a great data business. Thanks for tuning in and I hope you enjoy the episode.

Speaker C: It's always good to welcome back a friend of the show. And today's guest is Scott Brightner. He is the co founder and CEO, uh, of Kilo Code. They are an all in one agent engineering platform built solely for software developers. Scott, I actually met him four years ago when he was the founder of Brooklyn Data Company. He sold, exited, had 11 months out and then co founded Kilo in September of last year. They are a super interesting and valuable AI startup because they're operating right at the center of that shift towards agentic software development. The company is basically building an open source AI coding platform designed to fundamentally change how developers build software in the new AI era. Yeah, it's really good to have you back on the show, Scott. As I mentioned on my intro, you've obviously a returning guest, which is always good to welcome one of those.

Speaker A: Do something, right? I guess.

Speaker C: Yeah, exactly. And I, I, I was kind of alluding you've, you've had quite the journey, right? You kind of obviously founded, exited, and uh, a lot of people listening here think, right, that's the, that's the time to kind of put your feet up and kind of kick back, right, do some house runners, uh, spend some time with the family. But before you even the dust had settled, Kelo was born. So I'm really keen to touch on that because for people who haven't heard of Scott, uh, come across you guys, firstly, who are you? And secondly, Keelo is obviously super hot space. You've had hundreds of thousands of downloads, huge amount of interaction with your platform already. So yeah, really keen to touch on that. Firstly, who are you? And secondly, who are Kelo?

Speaker A: Cool, amazing, great questions. Straight to the point.

Speaker C: Yeah.

Speaker A: And excited to be back on. Um, so, yeah, so I'm Scott, I'm CEO, co founder of Kelo in a previous life and the reason Alex and I chatted, I don't know, you know, four years ago.

Speaker C: Four years ago, yeah.

Speaker A: Is that I was founder of Brooklyn Data, which was a data and analytics consultancy kind of early in the modern data stack. Build that up from kind of just working with startups and implementing data warehouses and data strategies for startups to working with some of the largest enterprise companies. And yes, I did sell it and I, you know, I did take, I did put my feet up a little bit. I aimed to put my feet up for a year. I made it, uh, exactly 11 months.

Speaker C: There you go.

Speaker A: That's enough. I know. I thought I did a great job. I mean I, I think, you know, my realization during that time was like I was nostalgically looking back at my journey from Brooklyn Data, uh, and just the excitement about being part of the modern data stack community and not just building the company, but like it felt like we built a community. I listen many, many, many people had far bigger impacts on the community. But I felt just deeply involved from the early days. From seeing these small companies organizing local meetups to, you know, they're kind of household names.

Speaker C: Yeah, yeah.

Speaker A: And I kind of nostalgically looked back at that and kind of as if right on cue, I got reconnected with uh, my good friend, uh, Emily, uh, Sherryo, who is my co founder, and then Sid Sabrandi, who is our third co founder. And we started kind of working on Kelo and I mean the kind of the story there is. And personally, um, it felt like the excitement of the modern data stack wave, but times a million, because this is AI. It's the opportunity to be thinking about how I know people are, about the future of, of how people are using AI to do their jobs, to do coding. I mean it's fascinating. And so like, I mean it was like heck, ah, yes. From the get go, it felt like all the warm fuzzies from the early days of Brooklyn Data all over again. So like, I was thrilled. Um, I mean, where Keelo plays it's, you know, I think we've got a little bit of a contrarian view, uh, on the market. Uh, you know, but I think what's really exciting me is that our contrarian view is really, I think it's proving out, uh, so we're a agentic coding, agentic engineering platform. Think, kind of writing code using AI. Uh, we're on VS, code CLI, uh, JetBrains, Cloud Code, Reviewer, Slack, you name it. I think the contrarian views that we have are, um, first, we believe you should be across all the surfaces. I mean we don't believe that there should be a crew of folks using CLI on one tool and idea another tool. Organizations are made up of folks that like all sorts of flavors. And so we should be bringing AI to you and not forcing you to come into a particular workflow or tool. And you shouldn't have to hop to another tool, just do code reviews.

Speaker B: Yeah.

Speaker A: The second is we're big believers that you should have the right model for the tasks. There are so many good models out there, not just the ones that are always making the headlines. And in fact we see the role of developers and to be honest, all knowledge workers as kind of being the conductors of a symphony of long running models from lots of different labs. And so at Kelo you can use models from any lab, local, stealth, open source, closed source, you name it. And then the third contrarian kind of take is we're not a big fan of the subscriptions where you know, use up a mysterious unknown amount of inference and then you get put in inference timeout for a few hours until you can use it again. And so that's us. You know, it's like all in one. Use any model and you never get throttled. Love that.

Speaker C: And it's kind of. I'll go back to your point about that warm fuzzy feeling about feeling, you know, I don't know whether it's you that's kind of um, you strike at the right time. Right. If you think back to kind of analytics engineering, when the DBT was taking place, all these companies were trying to, you know, analyst engineering was the hottest ticket in town. And now Kilo, where we're looking at certainly a lot of today's news and companies trying to make sense of how to build their engineering orgs. What do you put intuition down to? Is it just this kind of gut feel or is it just right person, right time and, and going with that?

Speaker A: Yeah, I mean, I, I think it's a combination of things. I mean I think it's probably 99.99% luck. So let's, you know, let's factor out the now we'll look at the 0.001%. Yeah, I think it's, it's, it's having a kind of a good networking community around you where you're kind of both learning about what's going on and you're being presented with opportunities. It is having a curiosity to dive into things that aren't fully baked and a tenacity to keep when it doesn't work right away. Uh, because you know, I think sometimes I'll talk to folks about like, hey, have you used AI? I mean for coding, but for like anything. And they're m. Like, nah, tried it, you know, six months ago, didn't work for that use case. I'm like, six months. Six months is a lifetime. If it doesn't work, check back a week later. You know, so you need the curiosity to try new things, but kind of the tenacity to like get that it doesn't work sometimes and keep going and keep trying and figuring it out because it's moving so quickly that if you write it off based on a first experience or like one bad taste or a rough edge in like any aspect of AI, there's a real chance that you kind of miss the opportunity. And so I mean that's, that's kind of my. And that's how I've kind of felt about it.

Speaker C: Yeah, that's fair. And obviously you're onto something in terms of this kind of agentic engineering piece. And you know, but obviously the idea, the, the excitement is one thing, but on your point is like, uh, the market's moving at such a rate. How did you know when you had that product market fit and how did you know it was that kind of moment when we could commercialize this? And I like your kind of contrarian view of actually how engineers should be exposed to multiple models and you know, on um, one platform they're not subscription based. But how did you look at this and, and ensure that A, the market needed it and B, commercialize that?

Speaker A: Yeah, I mean I think we started just with the general curiosity. So my co founder Sid, he's the one that first started kicking the tires on you know, open source coding solutions. I mean looked at, there's, there was some folks already out there. I mean Sid's background is, you know, founder of, of GitLab.

Speaker B: Yep.

Speaker A: So he's got deep ties and deep passion for, for open source. And so he was the one that, that first started kicking the tires just almost outta sheer curiosity. Like you know, Kelo started in March of 2015 and that was kind of uh, I think the models were good, but it wasn't like very, it wasn't obvious that like how quickly they would get to the level of capability that like that they are today.

Speaker C: Yeah.

Speaker A: But I think what was interesting is just like the momentum we and the we and the broader industry gained from there. Like I mean Kilo, I mean listen, we're not as household of a name of like Nike or Disney or something else, but I Have to sometimes remind myself that we launched 13 months ago and 13 months ago Kelo did not exist. Nobody knew us. And it's actually we have millions of downloads, um, so many people regularly using Kelo. We're now in the consideration set. What was kind of just so exciting for us to see even in our first few months, we launched in May, by June, July, we saw this just like uptick. And I mean I think it's just you ask what was the moment where we thought wow, this has legs. I think it's then that's when it evolved from, hey, this is a cool hobby project. I mean it was a company but it was like a project. We were a VS code extension to like man, we could build a really cool platform here. And so I think it was that summer of 2025 where it's just, we just seeing the momentum of demand and interest from our community was just, that was the moment.

Speaker C: Yeah, it's exciting, right? Because you touched on, I think you mentioned it like hobby. You know, you're kind of putting, dipping your toe back in and out of nowhere. Uh, you, you say that kind of 0.1% of luck, but actually it's that judgment and then capitalizing on that. And who's driving the, the adoption of this? Because obviously once you know, you've got this kind of reputation and you know, being open source, people are beginning to talk about it. Do you think it's from a, you know, engineer or uh, engineering managers heads off who are driving, trying to drive better productivity within their engineering org or is it actually engineers going in and saying we need to be using this and are you beginning to have you got a window on what that looks like right now?

Speaker A: Yeah, I would say like top down mandates on their own rarely work. Yeah, I mean I think, you know, don't be wrong. I mean like because you know you put a big poster and email round, it's like, hey, we need to use AI, uh, use more tokens. Like yeah, you are trying to a. You're trying to incentivize like metrics, quantifiable metrics, but they're not related to outcomes. It's like we need to be more AI and more tokens. That's, I like that's not related to outcomes. And then you can't just peanut butter AI onto an org and get the benefits. Because what we found is like interpersonal communication, you know, team structures. You really have to revisit all those things and listen, you can kind of get there over time. It doesn't need to be overnight. But like, you can't say, like, oh, we got AI now we're rocking and rolling. You have to really say, like, okay, cool, well, whereas we had, you know, two pizza teams. Is it like two pizza slice teams now? Is it one person as a team? Two people as a team? Can we be tighter? But on the flip side, I really do think that the success of AI is kind of driven by also grassroots adoption by individuals. Yeah, People wanting to do their job better, to want to kind of be more, you know, push the limits of what they're capable of doing, spend more time thinking about the hard stuff like architecture and vision, and less time about, like, where does the comma go? Yep, those are the people that I think are really driving things. And so, I mean, I think overall it's this great grassroots effort paired by thoughtful engineering and IT leadership that are just kind of understanding that it isn't as simple as, hey, let's use more tokens.

Speaker C: Yeah, it's super interesting because I've, I saw an article, I think it was yesterday that's come out around. We've seen this split happening in the market. Everyone talks about hiring market in and from an engineering sense. And you're seeing the top end of the market. So those engineers are adopting AI, embracing it, solving those architectural problems, becoming more efficient, freeing themselves up to do better work. Those people are, uh, inundated with opportunities. Comps have spiked. We're seeing that real demand. But on the people who have not so much less interested, but maybe not as focused on adopting it, we've seen them have become more. They're finding the market trickier. Are you beginning to see that as well? You're kind of, you're on the, on the train or you're being left behind?

Speaker A: Yeah, I mean, I think we're in this transition period. I think there was a period where if you didn't know how to use a search engine like Google, you could probably still get a job, or you didn't know how to use Microsoft Office, you could probably get a job. And then, you know, over time, and I don't think there was like a specific date in the 90s where it was true. But like, there was a moment where if you don't know how to use a search engine and you don't have Microsoft Office on your resume, you're not getting this job. So I think we're in that transition period where you can still have a successful career without being as kind of, you know, without bringing AI into your workflows. I think it will become increasingly challenging. And I think there will be a point in the next few years where if you are not competent with AI, and I get it, changes by sector and roles, it's totally not, you know, uniform. But there's going to be a moment in the next few years where if you're not bringing AI into your workflows and that doesn't mean like running five agents in parallel, but it's just like use AI to draft email responses and the presentation and help you do analysis and things like that. Like, I think that's going to be more and more critical. I don't think unless your job is extremely automatable and manual, will AI put you out of a job. But I think someone using AI better than you may potentially steal your job.

Speaker C: Yeah. So, uh, we're fortunate. We work with a couple of really interesting companies embedded. And one thing we've seen from some of the interview framework questions is the last couple of stages being very AI centric. How are you using that in your current workflows? And it's not about saying we don't want to hire you, it's just actually want to free you up to do more work or different work. So yeah, I definitely agree with that. I think it's a fascinating space. And how are you? Obviously you said that you went live in May, logged on in June, July, and you thought, wow, okay, something's happening here. Obviously the space is getting busy. There's more companies and products, uh, coming into this space. How do you keep pushing the envelope in terms of the product? You know, not so much a usp, but how do you ensure adoption remains high and that people, you know, still engage and use the platform?

Speaker A: Yeah, I mean I think it's, it's balancing the North Star of what you want to achieve. This kind of all in one agentic engineering platform, use any model. And I think like that's going to be what pays off. I mean you even already see pushback as some of the larger model labs have changed prices. People are starting to say like, whoa, I can't rely on one lab. That's a key dependency. If the lab changes the price or it goes down, the model goes down on a day. Does my 10,000 person team say, okay, no, we're not working today, everybody take the day off. You can't do that. And so, uh, our vision is starting to pay off and engineers are realizing that inference is a critical part of the infrastructure and they need to have a platform that's diversified. But I think that's the path that we're going on. Uh, and you know, I'm seeing that shift in the market, but we also need to make sure we're executing on like what AI looks like today. And that's, that's honestly a challenge. I mean, you know, we have to, we, we need to understand what is, is game changing and what is a side quest. And so I think that that's like, that's the big thing. Um, and I mean I think we have a really good sense of just kind of by being so involved in the community that it is by knowing what people there's genuine interest or what is just the press cycle. Yeah, but I mean it's every week you gotta read the news, you gotta understand, you gotta try the tool, you gotta talk to the users and says, is this what, you know, AI looks like this month and this year? Or is this kind of, you know, just uh, you know, the news cycle? Needs. Needs.

Speaker C: Yes. It's wild, isn't it? I think it's uh, every day you read a bit about what's the latest trend, what's one company does and you almost feel there's this kind of knee jerk reaction where one company does one thing. So, oh, we all need to be doing that. And then how would you describe in your own kind of, from all the work you've done, the current state of the market from an AI? Because I know if you think four years ago when we first spoke, who would have thought we'd be sitting here going, wow, things are changing so exponentially day to day. Data analytics engineering is a bit more uh, of a, it was still very new, but it was, it was a slower, uh, burn. How do you describe how cutting things are now?

Speaker A: I mean, I think there's, there is, it's, there are phenomenal advances every single day. I however think though that there's kind of the Instagram or LinkedIn version of life, that sometimes people are seeing that like, oh, wow, you know, you see these articles like I had my AI start a business and it makes $10,000 a day or like, you know, 99% of, you know, humans don't write code anymore. It's all AI. It's like, I think there's some really interesting things. And even on the killer side, like, you know, our team is, our engineering team has definitely shifted from writing code to more kind of supervising agents. But they're also getting their hands dirty. But I wouldn't, I don't think people should. I think there are some quite extreme claims out there that Might make everybody feel like they're extremely behind. But I will say that there's a real range of kind of experience levels and success. Um, there are people that are really on the extreme and using AI, and I think that is phenomenal. Kelo is one of those teams that, again, we've shifted to our engineers being. Instead of being teams of humans being. Being humans managing teams of agents. But I think the human is so, so incredibly important, and I don't think that's going to change anytime soon. Yeah, but I think there's real a range of adoption right now, from folks that have really embraced AI to folks that are new to it. And, you know, I think there are folks that are making proclamations of what the future is going to look like constantly. And it's making some people feel behind. Like, don't get me wrong, you shouldn't put your head in the sand and not get into AI. But I think it's very easy to assume that everybody has adopted it when there's kind of, you know, I think we're not even in the bottom of the first inning. This is so early. It is a fraction of a fraction of the impact of AI has been felt. And so that's maybe a zigzaggy way of saying, like, I think there's a real range of adoption out there. You need to be thoughtful about it. But don't believe everything you see in terms of the extreme proclamations that people said.

Speaker C: Uh, okay. I'm kind of glad you said it, because I think if you spend too much time online and read what said company's doing, then you do it can make you feel a bit despondent. And then. Yeah, there are also going to be some examples which we're seeing come out already, about companies making some expensive mistakes with AI. You know, kind of, they're thinking, I think you go back to what, right at the beginning, what you said, not every company's ready for AI, you know, and you can't just put AI on top of a lousy data environment. It's just not going to give you what you want. So I do think maybe we might level out and it just become a bit more normal at the moment. It just seems boom or bust at the time.

Speaker A: Yeah. But I do want to emphasize that everybody, like, if every company, every person, every role, you need to be using AI every single day.

Speaker B: Yeah.

Speaker C: 100%.

Speaker A: Just start somewhere. Just start somewhere. Um, that's the thing. It doesn't have to be like, you know, I have 10 agents running full time just Start somewhere and try to get better every day and you will be better than, you know, 50% of people out there.

Speaker C: Yeah, no, I couldn't agree more. And you're obviously an eternally kind of optimistic M and positive guy anyway. But, uh, people looking at this and listening, what's been the big challenge about scaling Carla? Because obviously with Keela, you got the May to June spike. You, you're obviously excited. Well, things are going well. Your engineering teams now focus more on the managing the agent side. But what's been the tricky part? What was kind of catching the founders up at night?

Speaker A: Yeah, but I think it's what we were talking about before, which is making sure we're on the pulse of AI but avoiding, um, side quests.

Speaker C: Yeah.

Speaker A: I also say it's like we want to be really thoughtful about how we work as a team. Um, how we are being very AI forward on how we work. Um, and there is no manual for that. I mean, that's the thing is. So we made a very conscious shift of actually shifting away from human centered teams to AI centered teams. Again at Kelo, typically one engineer owns a feature or kind of product on their own, and we tried to minimize the amount of collaboration with other humans. Humans are there to be talked to. And don't get me wrong, they could be your friends. But we don't want some sort of design document or PR to be waiting for two weeks while someone else reviews it, uh, for someone else to review it or to gather input from the whole organization. I think some of that is a little bit of a, kind of like a, like a safety blanket that people psychologically need.

Speaker C: Yeah.

Speaker A: And so I think we spend a lot of time thinking about what, how to work in an AI first world, how teams evolve, how to set goals, you know, the autonomy kind of control threshold. I mean, it just, those are some things that, you know, again, there isn't like a, uh, oh, I'm going to just go to the library and buy a book on this. Like, there's no book written. There's no book written. And if, if there was, it would be out of date by the time it was published.

Speaker C: Yeah. Actually that's something I've not necessarily thought about, is kind of the way of managing those teams and incentivizing, motivating them, promoting them is obviously very different. Now when ultimately they're managing AI, they're managing. Uh, you know, I, I don't want to use AI native, but obviously in terms of your organization, you're kind of practicing what you pre. Kind Of Kilo we are setting up the way we expect our, uh, users to set up.

Speaker A: Yeah, exactly a hundred percent. And so we've got this Kilo Speed manifesto that I think we wrote end of last year, um, which really kind of describes our view. Kilo Speed is just this effortless flow state that we try to keep our developers in. And it's one part having a great tool that's just works with you and lets you use any model, lets you kind of use any surface, lets you run multiple models at the same time. And the other is kind of the philosophies on how work is done. And so, like, I mean, I think every, there will be various flavors of this, but I do think, um, it's going to be just a really important area of thought and, you know, and, you know, organizations will have to take action on it. Like how teams, how teams are run in the world of AI.

Speaker C: Yeah, and that's the big thing, right? That's going to get that right. And you're going to have exponential growth, you can have roi, you're going to measure all that properly. Because at the moment it's very hard for companies to say, you know, what is the ROI on this? Are we saving time? Are we saving money? It's, it's a real tricky predicament in HR and kind of that. That's a big piece.

Speaker A: Yep, exactly. Yeah, I think it's, I mean, I think every big, every company is going to have to rewire big companies. Small, small companies will do it first, but big companies will too. Um, I think there will be, you know, pockets of skills or people that are kind of momentarily kind of. There's supply demand issues because of AI, but I think, generally speaking, I don't think jobs are going away. I think just people will be organized differently and they will be doing greater amounts of output per individual.

Speaker C: Yeah, 100%. And final question for you.

Speaker B: What, what's next?

Speaker C: Obviously, I know you guys are finger on the pole. There's a lot going on. Um, but what, what can people and engineers, who's listening to this, what can they expect to see from you guys?

Speaker A: Yeah, I, I think there is, there's lots of things we're excited about. Agents that are running increasingly autonomously for long run, long periods and sometimes even always on agents. How do you manage those folks? How do you integrate those agents? How do you integrate them into your workflow? Things like memory. There's a lot of interesting thinking around. How does your agent learn about you, your org, your project, your priorities? So that you don't have to keep saying, by the way, our. Our company color is green or like this. I mean, just like, I know that's a very simplistic answer, but like, how, how can it, how can it learn? And then just like continuing to, to, to make sure we're, we're, you know, leaning into this kind of multimodal vision of the world. Like, I do see this, like, you know, we're at this really interesting kind of fork where on one. One fork of the road is this kind of black and white dystopian world where there's like a handful of models that everyone uses. Uh, and on the other side is this, I think this beautiful Technicolor in Technicolor future where there's just hundreds of phenomenal models that do lots of different things differently. Some are fast, some are. Some are thoughtful, some are focused on science. I mean, like, who knows? And you're able to pick the right model for the job. And just for us, it's kind of leaning in. It's like, how can we do smart recommendations for which models to use? How can we enable kind of this great, um, M. Marketplace environment for new models? Just like, you'll see us leaning more into this kind of multimodal world.

Speaker C: Fascinating. There's no look, there's no, uh, there's no surprise. You guys have had this growth. I think that really kind of clear view, clear North Star, and actually appealing to the needs of your audience and that, that's ultimately is how you're going to keep this folk, right?

Speaker A: Yeah. 100%. Amen. Totally agree.

Speaker C: Scott. It's been great to have you back on. I feel we could chat forever, but no, it's been. I always kind of feel like come up a conversation with you, like, pumped. I'm like, right, we're ready.

Speaker A: Should we do it again in 2030?

Speaker C: I think we should. Uh, maybe I'll come to yours next time.

Speaker A: Okay. Perfect.

Speaker C: It's been really good to have you on.

Speaker A: Perfect. See ya. Cheers.

Speaker C: Scott.