Dana Lawson - Netlify

{/ TAB: SHOW NOTES /} This week we're joined by Dana Lawson, CTO at Netlify. We talk about her journey from the US Army to leading engineering teams at companies like GitHub, New Relic, and now Netlify. We discuss Netlify's evolution from JAMstack to AI-powered developer tools, including Agent Runners and their MCP server. We also explore the concept of "Agent Experience" (AX) as a new paradigm alongside UX and DX, and how hiring practices are evolving in the age of AI. - Netlify: https://www.netlify.com/ - Agent Experience Hub: https://www.netlify.com/agent-experience/ - agentexperience.ax: https://agentexperience.ax/ - Agent Runners: https://www.netlify.com/platform/agent-runners/ - Netlify MCP Server: https://docs.netlify.com/build/build-with-ai/netlify-mcp-server/ - Dana on LinkedIn: https://www.linkedin.com/in/dglawson/ - Dana's LeadDev Profile: https://leaddev.com/community/dana-lawson - Dana's UXDX Profile: https://uxdx.com/profile/dana-lawson/ {/ LINKS /} {/ Paste show notes /} {/ TAB: SECTIONS /} 00:00:00 Introduction 00:05:20 Transition to Leadership 00:10:03 Netlify's Evolution and Future 00:30:14 Architectural Practices and Intuitive Systems 00:32:57 Netlify's AI Offerings and Agent Runners 00:42:38 MCP Servers and Creating Seamless AX 00:45:37 Evolving Hiring Practices in the AI Era {/ TAB: TRANSCRIPT /} Dana: if you're prototyping, people are deploying. It's like local in the cloud. And so agent runners we're like, man, we know that people wanna use create. We know that they wanna collaborate. We know that they're using coding agents. What if we pull the coding agents closer to the deployment? [00:00:24] Introduction Andrew: Hello, welcome to Dev Tools fm. This is podcast about developer tools and the people who make 'em. I'm Andrew, and this is my co-host Justin. Justin: Hey everyone, uh, we're really excited to have Dana Lawson on with us. So Dana, you're the CTO at Netlify. Which is exciting. Uh, and you've worked on just, you've worked at a ton of incredible dev tools company, uh, we were just talking about GitHub and how you helped like ship the, the GitHub discussions product and then yeah. You've, you know, worked at Envision and New Relic and, uh, several others that I'm sure folks have heard about. Uh, you've been in the, the industry for a while to see like all the ups and downs, and we're really excited to have your perspective, um, before we. Dive in and, and chat more about, uh, you know, some of the things you worked about. I would just love to turn it over to you and like maybe, uh, cover any ground that I missed, like intro yourself to the audience. Yeah. Dana: Yeah. Well first of all, thanks for having me. Obviously I'm a Dev Tools fan, not just pocket. Ask, but also just in general. Um, you know, I've been doing dev tools like forever. It feels like being a practitioner and then being a leader. I, yeah, I had a really interesting kind of path into technology and I think my path is gonna become the common path now. Um, which is, which is, you know, not having this traditional development background. We've been talking about it for a long time, but I certainly didn't when I got into this rodeo, and it's been incredible to see. Oh man, incredible and scary to see all the cool shit that we're building these days and like how people with these unconventional backgrounds are also getting into technology. Um, so yeah, you know, it's been a, it's been a long road and I am, I'm probably a little bit obsessed with developer tooling. People are like, come do something else. And I'm like, but I have to help builders build. Um, and now the world can build. So it's been a really, a really fun time. Andrew: Yeah. Yeah. So let's start out with how you got into dev tools, because you didn't really start, uh, in the DevTool space and, uh, I think you have a little, little bit of an interesting journey. So can you fill us in on that? Dana: Oh man. Like, yeah, especially, you know, I'll just say it's, it's still intimidating to this day to get a meet like. It always sounds like I'm throwing some shade and shit, but like getting to meet like people with computer science degrees and like really, you know, somebody that has spent their, a lot of their life studying this, this science. Um, so I don't wanna make it sound that way, but like, I had no intention at all of doing that. Okay. I'm a weirdo. I wanted to do art. I mean, I always envision myself like painting for a living, and then you realize like, mm-hmm. If you wanna be poor forever, like try to be a, an artist, like straight up, like you gotta know some people from Eunice, New Mexico. Nobody even knows where that is. I don't even know where that is anymore. And so. I've always had a fascination though, and I've been a curious person my whole life. And sometime that's a fault. But in math and, you know, I was lucky to grow up through the age of computers, truly. Um, and you know, back when I was in high school, we were learning like Microsoft Front page and like HTML, like, like, like it was a new thing. The internet was like, it was new. New. And you know, I. I, I was being a rowdy kid in college and I was like, man, I don't wanna be poor forever. So I joined the US Army and they had this emerging field and like all I heard was, okay, I guess I'm going in the army 'cause I need something to do, but I really don't know if I wanna do army and computers need air conditioning. And like, I, you know, the little, the little like problem solver in me, which all we are, are problem solvers was like plan created army. Computers, air conditioning. Comfort and then future career and here we are, uh, because of that selfish nature of truly choosing my comfort. But in, in reality, you know, it was an interesting way that I got my foot in, when I joined the US Army, it was an emerging field, and I got taught for a year everything from protocols to application development in the late nineties. And then, you know, from then on, I, I got out of the military and really was a practitioner of, of dev tooling before DevOps. You know, we were all just a bunch of cis admins, um, and programmers that had to do cis admin work 'cause you know, stuff wasn't as separated. And I, you know, I've just fallen in love with it. And I think a lot of curiosity is what. What played into that ability to continue to excel in our field, because that's all it takes, man. It's like you gotta be a constant learner. And so that's, that was my journey. A little atypical. Didn't go to Stanford or MIT, like all my peers. Um, but if MIT wants to gimme a scholarship, call me. I'll go back to school. I'm, I'll go to school. I'm here for it. [00:05:20] Transition to Leadership Justin: So you've not only sort of had like an atypical journey in the beginning, but like you've been a leader at some of the like big name dev tools, companies that like a lot of folks. Know, and I'm curious about that transition from, okay, I'm a practitioner in the industry, you know, I've went through the army, I've got some training I've done, you know, cis admin or dev tools or whatever, like before the, the sort of industry that established around that to that path of like being a, a, an, a leader in the industry. That's like a, that in itself is like a non, you know, traditional growth path to it. I, I'm like curious, like, how did that come about? How did you work your way into like. Actually, what I want is like I love the technology and I wanna be a leader of people like in the space. Like what was that Dana: And I mean, there's honestly some luck in it, man. Like I think there's a lot of hard work and I always tell people don't ever, don't ever say you just got lucky 'cause you have to put work into, but there are some things that are outside of our control, but I think it's like having the awareness of when to seize those opportunities are really important as well. Um, and for me, I've just had a. I had a little bit of luck on my side, thank goodness. And I just really grind, man. You know, like I think that back in the nineties and early two thousands and, and you know, especially as I got into some of these bigger companies, you know, mid two thousands, the, the culture was a little bit different, good or bad. Like, I'm not here, I'm not, I don't even wanna go into that. It's a whole nother podcast for us. But it was more of a grind culture during that time, and there was this self-sacrifice. I'm not saying that's the key to success, but there is an aspect of like, you know, radical self alliance on what you wanna accomplish as a person that has to be put in to that ability to succeed. And then honestly, like, I like to think about it too, just like that stupid example I gave of like gamifying the army. But it's like, how are you, how are you projecting and seeing your future and what are you doing and what do you need to do? And then where do you find those opportunities? And it's like that in a thousand, like you're gonna, you're gonna get a lot of nos, but it's also being humble and, and starting from somewhere. And for me, like I didn't come into leadership even traditionally. I remember that moment I was. The traditional like tech lead. You know, you've grown from a practitioner where you're a senior engineer, you're a staff member, and you're like, cool. I'm the one that sits here and defines architecture or the systems in the paths and staff and above. You do that. Well, I did that so well and we didn't have a manager. You're just like, you keep telling everybody what to do. And I was like, I'm not trying to tell 'em what to do. I just think this is what we should do. Um, and I love people. Like, I think people bring me more joy than technology in some ways. So I was just like, always like rallying that team. I was like, we could do it. And I'm, you know, I'm like, I, I love what I do. I'm actually very passionate about it. And so for me it was just, you know, seizing those opportunities, speaking up. Um, and then also having good mentorship. You know, I, I luckily, again, stupidly I'll be like, can you help me? You know, people don't wanna ask for help. We see that as like a kink in our armor of success, and I'm like, no, go ask for help. Like, straight up, do it. I asked for help. I had a great mentor. Um, I continued to seek out advice from people that have done it longer than me. I think that's a huge part of my success. And then building real relationships with people 'cause they're the ones that are gonna open the doors for you, especially when the field is so saturated. It's like, who are the people that you can go and tap and be advisors, and then how do you grow your own self? Um, and I think that's a part of it as well, but. There was no magic formula. It was a lot of hustle, a lot of work, but a lot of relationship building and a lot of, you know, curiosity and humility, I think packed into it and then, and then hoping that the universe hears that and then opens up that timing. Like, you know, it, it was funny, like when I got the opportunity to join GitHub, I was on a panel with Jason Warner. Like that's how I met. That's how we met. That's how I got my GitHub job is like. We sat down on a panel, 'cause the universe put us together and we had a really good conversation. Then we started becoming homies and it was like, Hey man, you got the thing? And I was like, I, I do. And so, you know, that's how. Justin: Yeah. That's awesome to hear. And I, I think like your, your, your comment about like, it's not just about being lucky, but it's also recognizing opportunity when it comes along and then acting on that opportunity. Those are all like incredibly important things. Um, cool. [00:10:03] Netlify's Evolution and Future Justin: Let's talk a little bit about Netlify and the sort of like work that you're doing now. So, Netlify has had this like, very interesting trajectory. Like it was early in the space of when we were doing these, like static site hosting or, uh, what, what were they calling JAMstack? Uh, what was the, I forget like the terminology we were using. Yeah, yeah. Dana: That's . JAMstack. Justin: Yeah, absolutely. It was a really fun time where we were like, sort of like evolving these new primitives and, and I mean, Netlify itself has evolved a ton since then. Uh, can you walk us through like some of that evolution? Um, yeah. How did it get from like this JAMstack deployment site to this like robust like platform that it is today? Dana: I mean, I, you know, I'm very fortunate. Matt Billman is like one of those. What the genius engineers like the, I swear to God, he dreams about caching. Um, I'm glad he does. Somebody out there needs to continue to dream about caching for us all, but it was this really, this foundational idea. And I think, again, why I got excited about it is, man, it is so easy to over architect everything and lose sight of the problems at hand and what you're trying to solve. I've been working with the internet since the internet was a thing, and I'm like, man, we have made the internet hard, especially for like the majority of what it is, which is sharing information, buying, selling good trading goods, trading technology, trading information. Most of the internet is just that. There's some cool stuff that has to go behind it, depending upon the kind of business, but a lot of it's just, it's, it is what it is. It's easy, it's simple, it's small. And so you know, this JAMstack idea of really utilizing JavaScript. APIs and markdown was like, wait a minute. Like we can make most of the, the internet problems pretty simple, right? And then how can, and really what problems are you solving? Right? When we talk about dev tooling, it's like you're talking about time to market, getting the ideas out to the world immediately. And when you have traditional architectures that are full stack, which are still great, and you can still have JAMstack with full stack, it's just a separation of concerns so that you can move a little bit. Quicker and have this decoupling so that you're not bogged down by full system changes for quick changes. That should just happen, and it allows really this motion of iteration, right? Because that's all we wanna do is we wanna comment, like, subscribe, and change, change, change, go, go, go. It's everything. It's how humans are and we're only going faster. So I think that idea was really that decoupling and really to unlock engineering teams at the time. You know, those poor web developers, I'm a terrible web developer. I was the one on the other end being like, we'll, we'll get to it. You're like, yes, that's cute. You added a new banner. Like, we'll get to it. It's like, man, how frustrating is that? 'cause like in software development, man, there's some cynicism. Everybody thinks their job is more important than other persons. It ain't. So, you know, the JAMstack was really about like, how do we really enable that, that ability to change rapidly by separating concerns, allowing people that do have that ownership to do what they need to do and do it through this, this API driven architecture, right? And I think the evolution of that is like, wow, stuff with these agents now, like it good separated, even more stuff's going even faster. You're at computer speed now instead of human speed. And I think that the desire. For those serotonin hits of changes needs to happen quicker. But there's still a lot that happens under the hood. And so the evolution is continuing that same journey of like, how can I help people work? Which sounds like it's out of both side of your mouth independently, but collaboratively, collaboratively, quickly. And I think that's what really agents are doing for us. So to me it's like this workflow. The idea into production has just got accelerated, but it's still the same workflow. It's just now you're just like, you know, and, and because I've been a DevOps practitioner and dev tooling practitioner for so long, I'm like, you know, it's, it's just hyper speed of the CICD that we put in. But now it's done a little bit different and now you're putting, you know, the first part of that flywheel into some more. You know, air quotes, automation, and you're picking the best one out of those to send through the system. And so I think it's, it's, you know, it's, it's an interesting time and, um, yeah, I mean it's just, it's where I think all technology is kind of moving and that ability to move fast. So when we think about, again, those initial concepts. 10 years ago, JAMstack, it's, it's just about decoupling and composability and, you know, that Silicon Valley word future proofing, but it truly is. Um, and that's, that's again where like just having APIs, okay, like this started with the journey of like API endpoints and before APIs. There was other versions of that as well. But you know, it's been a long journey and now we're just on super speed. Andrew: So you just said yourself, like you're a, a bad web developer on other podcasts. You said you're a terrible JavaScript developer, but you're working at a front end company, like it's a billboard for front end, uh, people. How has that, like, has that surfaced any challenges for you and how has that shaped the product? Dana: I think it makes the product better because like, man, like, you know, I, I, I'm a, I'm a bad. Front end developer, but I'm an awesome systems developer and I know that I can make these systems amazing for y'all because I need to let you do what you do best. Like I, I'm trying to take away the crap you don't wanna think about. Like, I've been doing that for years. I'm like, oh man, you don't wanna have to go talk to them about, you know, the telemetry or the, the, the rum of your website. That's why I was at New Relic. I'm like, cool, front end person. Do it yourself. It's been about this enablement. Like, I think if you try to put a front end expert, I, which I, I can still do some okay. Dollar sign the world for life, whatever, but like, you don't want me in there doing that. We all gravitate to the things that we love and I love systems and I love scale. Um, and I think solving those are incredibly important. But the workflows of a JavaScript developer, y'all know better than me. I'm here to listen and take away all the stuff that you really should care about. Networking, actual protocols. Not MCP, I'm talking TCP ip. Okay. Stuff that like, you know, has to work with computers, um, ly. And so that's where I think like it, it is actually a good mix. 'cause I am surrounded by amazing web developers, right? And they're telling me their problems. Like a web developer doesn't typically work on a build system. Would you want a web leader building your build system? Maybe this day and age, I'd be like, hell yeah, do it. But I think you're probably gonna want, you're probably gonna want a different type of nerd, and you're want nerd that does that well, but you're gonna get the experience out of it. That's why there's still a ton of jobs to be done. And I think it's interesting that I am in this position because I'm looking at it through the lens of like, enablement. You know, you can't be good at everything. But the things I am really good at, like that's the thing I'm trying to solve for is like, how do we, how do we create. A golden path for y'all just to unlock your ideas. Like you have the ideas, like I'm just building the tools. Justin: I like that framing. So you're talking about enablement, which is, you know, why we're in dev toes enabling people. Um. And then sort of like to take this back, you're talking about like opening netlify up, not just to, you know, the traditional front end crowd, but like more to the world of like, now people have these agents and they can spin up these experiences and, and companies like Netlify are well positioned to sort of help people take these nascent ideas, um, that they might not have the full experience about and like actually make them into something real. And you'll have this, uh, you know. Term you've kind of coined, uh, which is like agent experience, um, as a, you know, sidecar to like developer experience or user experience. And, and I'm curious if you could just like, kinda, uh, talk a little bit about like, like where did the, you know, where did the framing of that come from? Um, and yeah, just like how are y'all talking about it internally? Thinking about it, like what does that look like inside of notify. Dana: Well, I think like, you know, the term developer. Is definitely shifting. I mean, it still exists, but like you go and read any of like the tech circles, it's like, I haven't written a line. Even the hardest core techies, like, they're like, I haven't written code for anything. What a what? What a dude from a Claude say today. He is like, I use Claude Red, Claude. I'm like, I know you do. Like I know. It's like nobody, no, there's still a huge conglomerate of. What I will say right now, we put, as a typical developer, that is still existing, right? They're outside of this AI hype, but I'm sorry, like it's come, this hype is real. Yeah. There's a bubble in it, but it's not going anywhere. We've unleashed the hounds. Okay. It's here. And so I think this expanding definition of developer is, is really coming into the people that are far beyond the practice of computer science. Are far beyond even like me in the, in the sense that I did have some education through US Army, but then hands on and it's now you're sitting behind a box that does it for you and you're just talking and it's incredibly powerful and it only gets better the more the people use it. That's how it works. Gets better the more you use it. And so when we think about this evolution, it was full stack dev. Now it's backend front end, now it's services, now it's DevOps. You're seeing a conglomeration of that persona come into one. And it's one, but it's wide. And I know we hear a lot about like taste and like all that matters, taste. And I agree it does. There's still jobs to be done before we get to that nirvana state of just good opinions are the only ones that count. Um, and there's still governance, privacy, and policy. And so when we talk about. This wider definition of developer, there is this component of citizen developer and that, and, and a portion of that citizen developer will become a professional developer because they're selling their idea and they've never written a line of code, but there's tools being built around it to make sure the checks and balances are there. And so when we talk about this agent experience, it's about seeing this evolution of how we think about the interaction. All this code. Right, and I think the easiest way to think about this is when we, we went and coined user experience, it was pretty clear that the output of the software is to do X for a human. The end user developer experience was this development software will do this for this builder. This agent experience is like your builder is now a robot. It's still dictated by you, but it can do things at scale. It can do things that we would do. It would take hours and hours for us to do or to set up. It can do it at scale. And so it's really, it's, it's about how do you, how do you apply the logic of what you currently serve to this new artificial user that can do stuff faster and can do it at scale. So agent experience isn't reinventing the wheel. It's not saying like, oh man, all that beautiful API calls that you had, um, you gotta go rework them. But it is going and saying, Hmm, now that you have an agent hitting that endpoint, you have your rate limiting, right? It, it could, before a person would hit a couple times, maybe you wrote a script and it would be like, great. And, but now you have some wild, indeterminate, crazy developer that's gonna go and pound your systems. Maybe you need the backend systems to be bigger, and maybe you need to rethink about who is the end user utilizing this. And there's always gonna be this like Venn diagram of that trifecta. Now, user experience, developer experience, agent experience, and the agent experience bleeds into both. Because depending upon if this is now your new artificial user hitting your site, or if this is your artificial dev building your site, there's still gonna be a entry point in. And so the way that we think about this is really about. Speaking deeper into, like, again, like in the back end of these systems and the way that we handle our tooling and the, the trading of information, what do we need to evolve and where does that, that lever of it becoming more ag agentic versus more human driven? And, and to me, the biggest delta, and there's probably some others, are again, scale and speed. Takes us a little bit longer, then it's gonna take the machine. But the experience is probably gonna be the same on the end. It's just, again, what are those systems in the back and how are they behaving? And, and then what matters. And of course, some bias here, right? I think about performance all day. So I'm, I'm sitting here going like, you know, what are the things that matter? Governance, security, privacy, performance, uptime, reliability. 'cause the experience. I have to trust that to be, again, for curated and created. But there's still the law of physics and there is still boundaries within computers. We don't have the quantum computing yet. We don't have the endless power sources, so I think there's still a big amount of optimization. So agent experience is really that in Twofolds is, is really updating, creating and curating for that, whether it be end users or internal developers and adding that layer. Experience on, on how, on how you approach building and how you refactor and how you rebuild. 'cause there are gonna be moments where you're gonna be like, it didn't serve. First of all, half our APIs barely serve our purpose that we designed for today. A lot of our craft 'cause like, you know, maintenance on APIs hard and it's like, okay great. What, how do we, how do we now do this when there's a million different variables in play that we don't even know? It's hard and so it's gonna be interesting and that, you know, there's a lot of talk of agent to agent and all of that, but this is again, where I think this evolution of agent experience is coming through. And again, that user base Andrew: Yeah, the user base changing is kind of wild. Like I've seen so many developers on Blue Sky that are like, I don't know code at all, and they're like producing apps that other people use. Like I literally saw this one person that's like, oh, there's this weird web pack error. I see. I hope, I hope I can push this production and it's okay. And it's like this totally non-technical person working. Like at the level we all worked before AI tools like. Effectively and providing support for that through like services is gonna be like a whole new set of challenges when like you're used to dealing with like technical people and now this, now the feedback you get is like, it don't work. I don't know why. Dana: Every, I love it. I love it because I love people and I tell myself this, but you know, you're like, oh man, I, I, we get a lot of questions on DNS, like, first of all, DNS is hard, okay? It's hard for us that've been dealing with this crap forever. You think of Be easy, it's a bunch of records. It is not. And so like you get a lot of, of these new citizen developers, they don't know what that is, shit. They do not know how the internet works. And so you're like, okay. Where. Where does the teaching moment matter? And so on this, I'm like, on your version, like they're probably gonna take that error. They're gonna go and chat GPT and be like, WTF, what does this mean? It's gonna be like, dude, they probably already puts it to production trust. I mean, I, you know, the security space is just gonna get better. CVEs, oof. There's gonna be all sorts of nonsense. Um, but then on the other hand, on like some of the like stuff you should, again, I say keep a beginner's mindset. Be curious. It's like. You can go ask it what DNS is and then you're not gonna know how to like fix dns. You're not gonna know how to fix DNS. And so again, I think there's a pretty big gap. But yeah, you read it on Blue Sky, you're like, oh my gosh. Like, and it's real. I see it every day. I see it every day. And they're really, on one hand you're like, heck yeah. This is incredible. Like people. Because we're the, we're the ones that are gonna get the experience out of it. Like who, who are we, the gatekeepers of saying, your idea is better than my idea. We're the, we're the unlocks of it. Like, I don't know. It could be anybody anywhere have the next, next best idea and Apple may fall off their head like Newton. I don't know. But like, it is, it is, it is wacky when you're like, man, I've spent my whole career being this person, and like, it don't, and they just caught up in 20 days. It's, it's, it's a little hard. It's like, man, that's why I'm like, maybe I should go into cosmetology. Maybe, you know, robots and scissors don't mix. Um, but no, it's, it's interesting. I think you have to look at it in that lens of, you know, we're gonna have to solve the bigger problems because of this. And it is true. Andrew: Yeah, I've, I've had to start looking at my own role differently. Also, like now that I can wield concepts rather than implement code, it like kind of changes how I view myself. Like in the past, like, oh, I wanna add vector, embedding search to whatever. I'm like, oh, that's not something that I do or wanna like spend weeks like getting the hang of. But it's like now I can just say, Hey, I know about this concept and I know about this tool. It kind of works for that concept. Go do it for me. And the same thing works with like platforms. So I've been like using railway to deploy a side project. I haven't interacted with it at all. I just use their MCP server and I talk to my chat and it sets all the services up for me. Dana: Yeah. And, and that's where again, I'm like, it's great to be a bad front end developer because I have the world's biggest helper. I'm like, like, everybody else has said, you know, I've written, I've, I've prompted more code than I have in a long, long time, and it's because I have the foundation like you do That says, oh. I, I know I want to try this out, but I'd have to go like practice and learn how to do it. Like, you know, I've been learning rust forever. Um, but now I can be like, I know I wanna use rust. I know how it would work. I know why. Cool. Go help me build this and that, and that. That's where it's like you're short cutting, but you still have that foundational understanding. And I feel like as leaders, that's a gap. And as a platform provider, it's a gap that we gotta fulfill. You know, it's like bridging those together until they don't matter anymore and they won't matter. One day, I'm sorry, they won't. Well, it'll be a, a different set of problems, but these ones that we're hemming and hawing about ain't gonna matter. Justin: One thing that I, I suspect like factors into the sort of age agent experience conversation that I've seen in my own way is like. Oftentimes building good tools, building good software for humans makes agents better. And you know, it's like, you know, semantic, consistent APIs can matter a lot. Really, really good error messages that like give a lot of context and detail about what went wrong and at least points you to some resources of here like what to do. Um. You know, in, in the past we have like been like, okay, well this is consumed by like a very technical person and like we're gonna give like a minimal explanation and they'll figure it out. And it's like, no, there's like a little bit more extra thinking. It's like, well, you know, we want our tools to be able to navigate this in some different way. And I like the knock on effect of like, actually as someone coming through who may be not as well versed in all of the systems. These new error messages are great. You know, it's like, it's actually giving people more context. And my hope is that as we build software to be consumed by agents in some ways, that we actually make it easier for people to consume. That's not necessarily a guarantee, but, um, I hope that as people are building things and thinking about agent interaction, that they we're actually pushing on like better experiences in general. Dana: go hand in hand. Exactly right. It's like if you're going to be doing this in. Human understood and readable formats, like guess what? You're gonna have to do it that way better as well, um, for the systems 'cause it's still all been learned off humans. And so like. It's, it's, it's, it's built off of us. [00:30:14] Architectural Practices and Intuitive Systems Dana: And, um, I, I agree and I think that's the interesting thing where it's like, where those intersections start happening is, you know, the, the, you know, the architectural practices and the best practice, just the best practices, architectural aside of like how to make stuff intuitive for machines and humans so that we can continue to ask the right questions. Get the right inputs and outputs in the system. And, and a lot of it is in actual language. And I do agree, and I think that if we do software better, it teaches the patterns of how to make better things as well. You know, there's still, we're still so far from it just being totally autonomous. So I do feel like, you know, you know, I was being, I was making a joke. I'm like, man, you know, like, how much do frameworks matter anymore? Like when, when Claude's, when I'm like building me a website, I'm like, whatever, Claude. Pick whatever website, um, you know, pick whatever framework works best. And so I think there's still this need though to teach the right use cases so that we can continue to get the best experiences. 'cause I've also had these agents build crazy overkill stuff. I was building a bingo app. Bingo. A fun dumb game. Okay. But I had some stupid rules in it. Nothing too nuts. And it's like, fire base this. I was like, man, that is so overkill for my Bingo app. I'm like, like how much money am I gonna, I'm gonna spin up me a whole fire base stack for some bingo. I think so. Um, so I think there's, well maybe I would, I did, I left it 'cause I was like, oh, this is interesting. What else are we gonna do? Um, it, it got a little overkill. It cost me too much money for Bingo. Um, but I think that, yeah, I think that's how systems are gonna get better. That's how we uplevel. But I think, you know what Andrew was saying, like all of our jobs are changing. My job's changing. Everybody's job's changing and you're being asked frankly to do more. I dunno how else to do. Like you're being asked to do more. That's at the end of the day. 'cause it's like, cool, you got a tool, go do more. And it's like, yes, but didn't we do this so that we could do more but not actually like continue to, to do more. Like, yes, I've optimized what took 40 hours. Now you don't want me to do 80 hours I produce. So that's where I think stuff's getting a little dangerous, but we're gonna, we're gonna, we're gonna recognize that, you know, yes, we all have to do more. And I think making it easier, um, is one, one part of that and making it more intuitive Justin: Yeah, a will. A little caveat that the production of more code does not necessarily mean shipping more code is, is faster or easier by any means. Um, I read so much more code now than I think I have in a long time. Um. Cool. [00:32:57] Netlify's AI Offerings and Agent Runners Justin: I, I would like to move on a little bit and talk, uh, about some of, um, netlify like AI offerings and on this podcast we have, we've sort of like struck this balance 'cause like so much of the industry is talking about this now and we, we haven't really dug too much, too deeply in a lot of the conversations with tools that are using ai. But I think the offerings that let Netlify have are really interesting and, and I think probably on the forefront of what we're gonna see from more. Companies, um, as it has been since the JAMstack days. Uh, so you have this, um, this feature called Agent Runners, um, which essentially lets you like prompt, like claude or, uh, codex or Gemini, and then do that deployment sort of directly. This like very, very fast feedback loop where instead of, it's almost like, uh, local in the cloud. Um, and. Dana: Yeah, basically. Justin: I'd love to just chat more about that. What was the inspiration for it? Like how do you see that used, like, et cetera. Dana: Oh man. So we had been trying to solve this problem for a hot minute, right? It's like we've made it super easy to connect a Git repo, do your build and deploy. You don't need know build system, you gotta do this. You're gonna just deploy. It's gonna feel like magic. And it's like, man, these cool coding platforms and agents like you are on fire, but there's still this handoff that's happening. We had this product called Netlify Create, which was really about this, this abstraction of a visual editor so that non-technical teammates could come and change the website and not have to go into the CMS and not have to go do all the nonsense or not have to have this web developers. So we had, we had this vision already of like, you know what, this can be easier. And so we, we put out this, this visual editor. It was hard. Okay. Like you, you gotta do a lot of work on the undersea, the like aligned components. It was, it was, it was really hard. Right. But the idea was right. Okay. Like we knew, and we still know that everybody wants to participate. Everybody has a dang opinion. You have an opinion and you wanna iterate. Faf an opinion. And now these coding agents, man, you're on fired. Like, you're like, change, change, change. And what we see is this behavior. Of Netlify usage already is like if you're prototyping, people are deploying. It's like you could see it local dev, but like you said, it's like local in the cloud. And so agent runners we're like, man, we know that people wanna use create. We know that we want, they wanna collaborate. We know that they're using coding agents. What if we pull the coding agents closer to the deployment? Because really people are not. Going and sharing like their 1 27 address and then saying, here, go install my website. Tell me what you think. Like, don't wait. It's a pain in the butt. Like to go download. I mean, it's a pain in the butt, but it is so it like when you have a stakeholder that's not working that way, and when I was at GitHub rollback time, you know we did code spaces. Which was code Spaces was meant to be your in the cloud. It costs a lot of money. It was still just for developers and you really couldn't have your, your stakeholders look at your stuff. So it was good for a purpose. So, long story longer we're like, man, we still think it, it'd be great for rapid in innovation. And the site's already deployed. The build's already there. We already have preview deploys. We already have this concept of branch deploys. Really all we've done is we've pulled in that agent experience into Netlify and we're allowing you to use that netlify flow of being like, cool, here's the website. You wanna go Say, great. I have a new campaign, I have a new banner. We, we make agent runs for everything. I think the design team are gonna smack me if I change the navigation bar. Don't worry, I haven't merged anything if they're listening. But what you can do is you can go and experiment and run stuff and it does it on a protected branch is a live preview. And now you can hit that agent run against Codex, Gemini, et cetera. And you can see the different ways that these models and agents are gonna approach it, which people go do it. We're the AI gateway, like, don't even worry about it. You know, you're running this through PHI so you don't have to go put in your credit cards, et cetera. So it's like, great. I wanna go make a rapid change on the website. It's already live in production. I've come in here, I said, Jim and I run this, and I'm looking at these. Three of them have a live preview. As soon as I'm like, this one is the right one, you like it too, I can send you the URL right now. I hit merged and it's done. It's done. You're still gonna go through your CID pipeline, so if you've configured any checks and balances, that's still gonna happen, but it just makes it quick. And yes, you could go into cursor Windsurf BS code, but you're sitting there having to say, deploy, check, and the workflow brings you out. What we find is once you deploy the first time. You can iterate right there. And it is a little bit different for traditional developers 'cause we're used to working in our IDE, but when you're but not as new hot, and I'm calling myself one now, not as new hot frontend devs, we're just like, sweet here, go do an agent, run yourself. Maybe you don't merge, but you can show your stakeholders and run and just makes that feedback loop incredibly quick. That, that process frictionless. And I do think you're gonna see more and more of that. Um, why wouldn't you? Because I always said, and we've always said, there's no better testing than in production. And it also changes the way you think about AB testing. Now with this, you can go really take it off to the res and say, I wanna do some double testing. 'cause at PHI again, we have, uh, a, uh. An intelligent network that utilizes functions so that you can display what you want and those different experiences. And there's a lot of ways that you can get that feedback very safely because your controls and governance are still there. It's just now you're just rapidly iterating. And it's been pretty fun. It's been pretty fun. I said, you know, the one thing I'll say is like, um, you know, we're working on really trying to think about the different people that are gonna use that because they're gonna wanna have different controls. Like I was joking. Changing the navigation of a a website is a big deal. People don't go do it in, if you have a design team, go talk to 'em first. But like before, that would not be something that you're just gonna go and approach. Uhuh, Uhuh, uh, Uhuh. You'd be like hands off or maybe, you know, um, there could be, you know, you have some old monolith and you're like, shh. I don't know. I'm gonna change this one thing. I don't know if I'm gonna change another. Perfect. It's like give you some safety, 'cause it's gonna scan that monolith faster than you can. That's again, where I'm like, use an agent. Like you're not gonna go read all that code. I mean, I'm reading way more code. But that's really what agent runners is intended to do. And there's other things that you can do beyond that. Like I like to kick off with, run on the same prompt 'cause I wanna see the difference, but I can also serialize my work streams. Kick off one prompt here, kick another one off here. And then I'm running three fixes at once. Because you're all get based, guess what? You're not gonna have merge conflicts. It's like you are now three developers and it's pretty freaking sweet. So now I'm a really good front end developer. I don't have any taste though, so I'm not that good. But I can't sling the code. Andrew: Yeah, that, that also must just open up so much opportunity for other non-developers to contribute to the process. Like especially since like with some, some of these platforms where it's like it prompt to get in apt you, you like they hide the code completely. And as a developer myself, I'm like, well, sometimes I need to go in there and like look at what you're doing wrong so I can prompt you to do the right thing. But having that collaboration and that shared surface seems very powerful. Dana: Yeah, because I can go see, I'll be like, I, I, I still think it's great for, you know, um. Tech writers, content people, it's like, cool, you're not going in CMS, you're not going anywhere. It's like, go like, and us as a developer that still have the keys to the queendom, I'm like, great, I can see what changed. I gotta make sure it's good. And I would much rather move faster on that than like waiting for that ticket to come in, fill in that dang ticket out. Especially when it's a change like that. You just wanna make it go. But obviously you talk about all day. It's a pretty cool tool. Y'all should try it. Let me know. Gimme your feedback. Justin: Yeah. I like the idea that, you know, there's so often that folks will just have ideas, but like not have the expertise. And I've seen companies, and mine has experimented with like giving a lot of. People across the organization access to like cloud code and other tools to like, yeah, you can pull down our product and like play with it or whatever. Um, but environments are hard to sync up. You know, you like got all these environment variables and all this stuff that you gotta do and you gotta worry about security and you know, yada, yada, yada. And oftentimes it's like, no, all they want is that experience. It's like, all I want do is like have a magic box where I can ask it to do a thing and see the thing and then show somebody else and that's it, It's great. Like the chief revenue officer was like registered agent Ron. On the website. 'cause he wanted to add some documentation for something he was doing. And I was like, all right, lemme see, lemme, I said, let, let's lemme show you some pair programming. So I went and I looked at it and I was like, exactly to your point, Justin, before he would've had to go have, get installed, install the, the, the, the Netlify app locally, which, good luck homie. Dana: We're a 10-year-old company. And then try to figure out how he's even gonna get that coding agents to do anything like straight up. He could just be like, cool, I wanna add this documentation. And it's, it just, it's there. I'm like, dude, you just added some documentation. Of course I'm gonna merge that boom. Um, but still teaching him that practices of having a PR reviewer and looking at it, it was pretty, pretty interesting. But yeah, it's, it's, it really takes out a lot of that setup, which, you know, it's scary, but also, but also scary. Andrew: Um, you so. Justin: next one. Andrew: Okay. Uh, so before we move on to our closing questions, uh, let's talk a little bit more about your AI offerings. [00:42:38] MCP Servers and Creating Seamless AX Andrew: Uh, you mentioned AI gateway, that seems interesting and can save you a lot of money, but lately my mind's been on MCP servers. Uh, I work on a design system here at Snowflake in the MCP server was just like. An enormous unlock in like productivity and like correctness within our system. And I was like kind of blown away by how, how much that needle had moved in about a year. Uh, you guys have an MCP server too at Net Netlify. Can you walk us through like what it does and kind of what it enables what. Dana: I mean, just like I, you know, you gotta think about this, right? Your MCP is your UI to your agent. It's just giving you the boundaries of what it can and cannot do. So our MCP server does all the things that you expect to get from the. Amplify from that CLI experience too. But again, so your agent can do it. And it's the same concepts that you think of applying that control system with your design system. I love design systems. I'm glad we're gonna talk about that. Um, but I think it is that huge unlock because you're given the written rule set of how to interact with the service. But for, again, an agent experience, I mean, CPS are the protocol for agents. Again, it's the ui. And so for us, we have a pretty simple easy M-M-M-C-P experience so that you can get all of this out of Netlify. You can go into your id, but the, the Netlify MCP, and then you can do all this stuff. I was trying to talk about and more, but do it in a more automated workflow. And again, it's to allow everybody else to participate with the right boundaries that you have. And so it's just another way for people to participate. Well, PE people, agents. Agents and the people us as well to participate with Netlify, but it mimics all of our other offerings. Um, but yeah, I think it's, it's really, it's really allowed you that sense of like. I've put the, the rules in place for the system to only do this one behavior, um, but be able to unlock that through that a, that agent, um, workflow. And so, uh, you know, there's nothing really fancy there. I don't know. Aren't MCP so 2005, like what, Andrew, what's the new hotness like? I mean, give me, gimme the hot deeds, right? I thought we were going back to just API or. Justin: Skills are the big thing now, I guess. Uh, Oh man. Skills are, you're right. we're going back to markdown, which like brings us back to jam, so, you know, uh. Andrew: It's always been about mark down. Dana: it's always been about that, about markdown, but why? Why are we not surprised that we're coming back to markdown again? Human readable, natural language marked down human readable, natural language. It makes freaking sense. I can't wait till just everything talks in, like, you know, new vernaculars. Then we, then we know we've really conquered all these when it's just saying, no bra, it's wrong. Um, so we're getting close to that. We're getting close to that, that, that landscape. But no, happy to talk more about our, our MCP, but yeah, that's exactly, it's pretty simple. You can deploy, you can roll back, um, you can set your domain, you can do all the stuff. Just gives you the rules of identify. Justin: Cool. [00:45:37] Evolving Hiring Practices in the AI Era Justin: Before we wrap up, I also want to ask a, a little bit of a divergent question in the sort of like leadership space. So, um, you know, as a, an executive, a leader in, in, uh, companies like Netlify, you've obviously had to do a lot of team building, a lot of hiring, um, and I mean, I think. Hiring in the engineering space is, is a fraught thing. We, we kind of all know this, everybody's been through the whiteboard or the live coding interview or, or whatever. That's felt like not great. And it's hard to, you know, in an hour or three hours or whatever, get a good sense of who someone is and how they'll work and whatever else. So it's hard for both sides. Um, and I'm sure you have a lot of, uh, opinions about hiring. Uh, I would love to, to sort of like just get your perspective on like building software teams and, you know, now sometimes we're in an interview and we're actually interviewing three gremlins in a trench coat and, you know, it's like hard to know that, you know, someone is. Who they say they are and, you know, they're actually representing their real knowledge and not, you know, uh, a live AI agent that's like feeding them answers or whatever. So just, yeah. How do you think about hiring and how has that evolved in the age of, you know, uh, intelligence on demand? Dana: Oh man, this is a tough one because it's like if you are part of this AI wave. You're sitting here going like, somebody is like gaming our interview 'cause they're using ai. I'm like, don't you kind of want them to, I'm sorry. Like don't you want them to, why wouldn't you want them to? Because you're gonna have them use AI at work. And if they get through an interview, obviously I don't want them to, but I'm saying like, you gotta take that with a little bit of like shit. Like I want their workflow to be driven through agents because I don't expect them to work that slow, but. On the other hand, I get three hours with this person. Like, do they know anything? Because it's, it's a balance. And this is the same conversation that we're having about like, you know, development as a practice today is like, where, where does, where does junior engineers, how do they get to staff when they're just sitting there and prompting, okay, you gotta change how you interview, you've got to change it. Whiteboarding is not gonna, does whiteboarding matter to your business? Yes. No. Maybe so. Probably not. Okay. What matters? Effective communication. Integrity, you know, like authenticity, passion, curiosity, and problem solving. And like I used to get silly interview questions back in the day that they do not do anymore to y'all. But it would be very philosophical, like. You have a, you have like an eraser and a pencil. Like what would you do with it? Why? You know, you're like, oh my gosh, why were you asking me a pencil and erase? What does that have to do with heck? But I do think that the way that you interview has to change because you're now mining more so for critical thinking than anything else. And I think that if you have something. That can be gamified or, or taken off the internet or utilize an agent, you're not gonna get what you want. You're gonna get exactly what you need. And it's like, okay. And you have to think about that whole experience. Like you do wanna see the use of, um, how do I utilize an agent to do my work? Like, do that in real time, like screen share, you know? But you have to change the way that you're interview and you also have to change your hangups of it. Like, I heard a whole bunch of people, like, I'm on a. I'm on a group with a bunch of CTOs, and this was like probably a year ago. They're like, dude, what do you think of like using AI in interviews? No way, man. Like that's how you're gonna suss out. And I'm like, y'all are all telling all your staff and now you all have mandates to your staff to use ai. Like how are you my like, do you wanna have a whole bunch of cynics come into your organization? Because their first experience was like, no, we don't want you to use any ai. We just want you to build it. Sorry. Like you gotta kind of embrace it. So for me, when I think about teams, there's two parts here, I think. What are, what are we evolving to, again, like this grounding in like our workflows. What do I want people to be great at? Again, I want great people and great people have behavioral traits that make them awesome to work with. And those are things that you cannot teach. You cannot teach somebody to be curious. You just have to be. You can't teach, you can't teach somebody to have integrity. They just have to. Um, so I think it's really about understanding the gaps within your team, what you're forming, and then having real ways to mine how they have that critical thinking when applied to building. And I think that pair programming is still the most, even though it may not be programming per se, it's still the most powerful thing you can do. And it's that concept of like, great, let's go open up your IDE. Shoot. Let's go up, open up anything. Let's just talk about it. Let's go build this app together. You and me. That's the best interview I could think of is being like, Andrew, Justin, we're building an e-commerce app. Let's go. Let's go bring a terminal up. What do you wanna ask it? What do you wanna ask it? Why'd you ask it that? What would you do next? Can't gamify that. Can't fake that out. You can have a little fun and see how they work well with others, see how they communicate, and then ask different pointed questions about the pieces of the loop. Especially as we were talking about this developer persona becoming a lot wider and the hats that we wear becoming many, and the responsibilities that we shepherd are behind, beyond building code and systems. Those are the things that you have to surface for because you're evolving this team. And the way that I built teams two years ago, I would not build them the way that I would build them now. I think there's some changes in the, in the approach that we have with, uh, with the makeup of the team. And really a lot of these tooling and the tooling that we've been building in dev tools forever is for autonomy. And so like you wanna build autonomous teams. You want 'em to feel connected to the mission and like having you be allergic in the A, in AI in the interview. Kind of like anti mission in a lot of ways. So I don't have a perfect rule for it, but I just try to look at, look at things more holistically and really try to break out some of those systems that we had in the past. Like, let me pass you this GitHub repo and go do something. Ain't gonna tell you shit about anybody. Or even looking at the contribution margin. Whiteboarding exercise maybe. But do you need them to do that? Are you gonna have somebody that's gonna hate working for you? 'cause they think every project should start with an architectural discussion. That's also something you probably don't want. You gotta think about that more. Well-rounded in my opinion. Andrew: Have some of those learnings made it way, made their way into the netlify hiring process yet. Dana: Oh yeah. I mean, I, it's, you know, we are, oh man, man, we are dog food everywhere. Our hr, um, talent coordinator, she wrote a whole new interview application. She vibe coded a whole new app and flow, you know, like you, she gonna vibe code. You're gonna vibe code with her before you get in because that's what she can do. Yes, we have changed a lot of the things, but we still do some of the traditional, traditional stuff. I keep a lot of the autonomy with the team, um, because, you know, they're gonna mine for different areas, but one of them being incredibly important is that use and that comfort of utilizing agents and that demonstration of how you would approach it in today's software, you know, workflows of how to use agents. That's the biggest part that shifted is really drilling into that versus just do, do, do computer. Andrew: It is wise. It's a, it's a changing landscape. So, uh, as our last question, uh, for someone who's been a CTO for such a long time at this point, what is like the one piece of advice you'd give to a person who wants to end up there in their career? Dana: Don't do it. No, I'm joking. Do it. Totally do it, do it, do it, do it. I, I, I think my one advice is like, you know. Figure out the, the, the path of your CTO version. There's so many, just like developers, there's so many different versions of being a, a chief technology officer and there's not one right way to be one. Um, and I think it's really about mining where you're passionate about that and, and navigating your career with a place that supports that version of your. CTO. Some CTOs are really just chief architects and they, they don't do the leadership and business side. Some just do the business side. It doesn't matter. And so my advice is, at the end of the day, go for it. Just go for it. Grind. Stay hungry, be curious, and make relationships. Um, because anything beyond, like, you know, when you're talking about the leadership rungs. It's, if you're not hands to keyboard, it's all about relationship building. But you have to be smart too. So like, my advice is stay curious, keep, keep prompting code. You ain't writing it. Um, I don't know. And, and I don't know, go for it. Like a lot of people don't think they can do it, but you can do it. If I can do it, you can do it. Justin: Why is advice? Andrew, you're muted. Andrew: Uh, well, that's a great note to end on. Thanks for coming on the podcast, Dana. This was a, a lovely conversation about all things AI and Netlify. So thanks for coming on and talking about it. Dana: Thanks for having me. It was fun. Justin: Yeah. Thank you so much. Loved hearing your experience. Uh, yeah, I don't know, need to check out some of new net's offerings. And you know, also you mentioned earlier you're a 10-year-old company props for continuing to innovate. 'cause like sometimes, you know, it can get stale and you are like still pushing at the forefront and it's exciting to see. Dana: Aw, thanks. Thanks for the love, and I hope people go check it out too.