Event Replay: Stack Overflow & Learning to Code in the Age of AI

[00:00:00] Hey, everyone! [00:00:06] I'm Natalie Cone, your OpenAI Forum community architect and member of the Global Affairs team. [00:00:12] Welcome to the OpenAI Forum! [00:00:14] We're here today because we all see it. Software engineering is evolving faster than at any point in our lifetime. [00:00:21] The tools are changing, the workflows are changing, and the people who learn to adapt and lead through this moment are going to define the next era of innovation. [00:00:29] And the OpenAI Forum exists exactly for moments like this, where we can bring together the people building the future, share what's actually happening in real workflows, and learn directly from the people shaping the field. [00:00:41] So today, we're incredibly fortunate to be joined by two leaders who are right at the center of how software work is changing. [00:00:46] Prashanth Chandrasekar, CEO of StackOverflow, representing the world's largest developer community, and the person with one of the clearest views into how engineers are learning, collaborating, and adapting in real time. [00:01:02] And Romain Huet, OpenAI's Head of Developer Experience, who spends his days working alongside builders everywhere, helping them translate ideas into production-level systems using AI. [00:01:15] Together, they're going to dig into what this transition actually looks like, why AI is not replacing software engineers, but is instead reshaping the role, why there's still tremendous demand for engineering talent, especially engineers who can guide, evaluate, and collaborate with models. [00:01:32] We'll also get an early look at insights from StackOverflow's 2025 developer survey, and hear about new collaborations between OpenAI and StackOverflow that will expand what developers can do. [00:01:43] So, with that, thank you all for being here, and please join me in welcoming Prashanth and Romain to the stage.

[00:01:51] Thank you so much, Natalie, for the warm welcome, and hi Prashanth, thank you so much for being here. [00:01:56] Hey Romain, thank you so much, and yeah, thank you again also to Natalie for having me on this. Excited to talk to your audience. [00:02:05] Absolutely. We are delighted to have you. [00:02:07] I've been a user of StackOverflow for, like, I can't even remember how many years, like, since I got online, so really, really thrilled to have you. [00:02:14] I thought maybe to kick us off for people who don't know you just yet, like, why don't we take a little bit of a trip down the memory lane and tell us about it, like, how did you first learn to program? [00:02:21] Like, what tools and habits kind of, like, got you to where you are today?

[00:02:29] Yeah, no, sure. So I probably started learning to program maybe in 1995, something in that sort of, you know, general sort of zip code. [00:02:35] And yeah, I grew up in India studying, you know, computer science as part of, you know, high school education. [00:02:45] And my mother was a, or is, in fact, still a medical doctor. [00:02:49] So, you know, she's got a couple of specialties, but back then, she used to have a clinic, a medical clinic. [00:02:54] And one of the things that she was doing is obviously managing a lot of patients coming in and she was treating them, etc. [00:03:02] And I got inspired as part of, you know, some of the coding, the code that I was writing to actually write a hospital management system. [00:03:08] So almost like an electronic medical record, equivalent of today, you know, there's so many companies in that space. [00:03:13] And I wrote it, I think it was actually in C++. If I'm remembering correctly, it wasn't in BASIC, it was actually C++. [00:03:21] And I really just remember this experience of being able to generate this visual element. [00:03:27] And it was so gratifying to actually create this, obviously there's like a backend, you know, keeping track of, you know, records, but this beautiful interface of experiencing, you know, the interface of how you would interact as you enter this information and actually get out. [00:03:40] So that was just a, it was just a joy to do that.

[00:03:44] And that I think inspired my, you know, kind of my education in college as an example in computer engineering, which is sort of has computer science and electrical engineering. [00:03:53] And I worked as a software developer at a couple of large companies in a national semiconductor, so writing Perl code for the semiconductor fabrication unit. [00:04:04] So that you can actually, it's kind of an older version of pager duty of, you know, today's world. [00:04:09] So it actually send alerts to people who were sleeping. If something happened in the manufacturing facility and the, you know, the wafer was actually stuck in the process as an example. [00:04:17] So the old school pager that people used to carry around on their, on their, you know, their belts. [00:04:21] So that was, to all those, I think were very like real experiences to say, you know, how do you really make a software and make a difference in real life? [00:04:28] And I think that was quite exciting.

[00:04:30] Then I think I, I, then I realized that my interest and passion was a lot, a lot about like how these tech companies were scaling and, you know, how do they operate at that level? [00:04:38] And so I went more on a business career over time until recently, I just say Romain when I picked it back up, thanks to all of your, you know, Open AI and others who've made coding now again, a joy and know a lot more accessible. [00:04:51] So I actually started coding recently. I wrote a couple new applications. [00:04:55] One is the finance based on a healthcare based application, which I was very excited to kind of do recently.

[00:04:58] Speaker 1: which I was very excited to do recently, so that was fun. [00:05:00] Speaker 2: That's awesome. [00:05:02] Speaker 1: I mean, when you started writing code, I'm sure you didn't even have Stack Overflow itself, and every developer on the planet knows Stack Overflow, and I'm sure that must have been a very different experience from what you have today, with these tools, where every side project that usually felt like a big deal to even start working on them, now every idea you have, you can actually kick them off with Codex and these tools to actually start building. [00:05:26] Speaker 2: Yeah, it's like, you know, it's like, each of these are like abstractions, obviously, and since you've been a Stack Overflow community user, you understand what it was like to go from, you know, looking at your textbooks to just figuring it out yourself, to talking to your friends, to talking to maybe your teachers or professors, and then you add Stack Overflow, which abstracted all this knowledge in a very democratized way, so you could have the world's community members, we have about 100 billion people on our site, help each other out now in all sorts of new form factors, not only Q&A, but you know, chat, and even opinion-based questions now. [00:06:03] Speaker 1: And that was a big abstraction layer when Stack was formed in 2008. And then of course, when you folks, you know, launched your product in late 2022, it was absolutely another major abstraction, really kind of a major seed change in terms of progress in software development, because I don't think it actually has made a lot of definite progress. You remember the days of object-oriented programming, et cetera, but they didn't really sort of make a big deep frog of a kind of a difference. And now I think with what you folks have obviously pioneered, I think it's been amazing to get back into it. And I'm curious actually to hear from you. Like you've seen Stack Overflow evolve, like a long, all of these major shifts in technology and platforms. How would you say this like AI era compared to the previous inflection points that you've been through? [00:06:52] Speaker 2: Yeah, I mean, you know, like you, I think we've all been through the, you know, the internet becoming sort of mainstream. In fact, when I moved to the States, the way I was applying for college and I remember sending like handwritten notes. You know, that's what, that's what I did, no, it's so different from like when I do ask the colleges about, you know, scholarships and those sort of things. But the internet just more broadly and then of course mobile, which all of us experienced that we're completely sort of more dependent on it. And then of course we had cloud. And so when you think about the kind of, all of those are pretty large platform shifts, no doubt. And, you know, I think it would be understating, you know, generative AI as a platform shift. I mean, it certainly is a platform shift, a major one. [00:07:34] Speaker 1: But I think it's more akin to, you know, probably closer to the internet because I think it's a new way of doing things. And, you know, of course it's redefining sort of all aspects, all workflows need to be redefined, especially now we have a big enterprise business. We hear about this, about Stack Overflow inside companies and how they're leveraging our products. So it's a lot of, it's quite fascinating to see that sort of entire processes are gonna be completely redefined as a result of now having access to this amazing technology. So I think it's very fun. And it's not just like a new set of primitives and APIs and capabilities for your own products and apps, it's also like changing what it kind of means to be a developer every day. [00:08:14] Speaker 2: You know, like the whole definition of being a developer is changing. And you know, many people, including parents of aspiring engineers, kind of sometimes worry that AI will replace developer jobs. And I'm curious, like how do you see the talent landscape evolving? And how do you see the role of the developer evolving? [00:08:32] Speaker 1: Yeah, I think it's an excellent observation and question. I mean, the observation that you're making, I think is spot on. Is I think the way in which you write software has completely changed as a result of this innovation. I mean, when you think about the infinite number of libraries and, you know, scripting languages and programming languages, now with this level of abstraction, I think what we're really doing is trying to manipulate the system. I'm talking about the coding agents to be able to produce what you want to through obviously, you know, by way of skillful prompting. [00:09:02] Speaker 2: And you know, if anybody who's actually tried to do this, it's a completely different mindset to be able to run it with multiple agents or, you know, be accomplishing sort of certain outcomes. And so going to your second, you know, kind of your question, I think with younger people, if I, you know, I have a 16-year-old and a 13-year-old. And so I understand this question deeply in that it's, I think, very important for young people to have a combination of learning about the fundamentals on how these things are actually, you know, how do you write software well? You know, how do you get on this plane of becoming or, you know, gaining mastery on a subject? [00:09:40] Speaker 1: Because, you know, we all need to be good at something. So the foundation of what you're building, I think, needs to be still set. So how do you architect an application well? You know, how do you design it well? You know, how do you write great unit tests? You know, what questions should you be asking? And then of course, how do you build it?

[00:09:56] And then of course how do you build it to scale in a broader company context.

[00:10:00] And I think all those topics is traditional, let's call it, education around software development. So I think you, so my advice would be, I think for young people, to learn all those fundamentals while absolutely learning how to code using the latest AI tools. Because if you don't do that, then you're gonna get left behind. So it's a combination of both because at some point you're gonna be in a situation where it's on you to build something very important for a company. You better know what's underneath the hood before you push it into production. And, but at the same time, you can go fast with that combo.

[00:10:36] Totally agree. And I think like double-clinging on that, I think one of the things that you mentioned recently that I really enjoyed was the idea of demand for code is literally infinite. And I very much agree with that, but I'd love to hear you unpack that in your own terms. Like, what do you mean by that?

[00:10:51] Yeah. I mean, if you think about, you know, I'm a big, I'm a big Star Trek fan. I've always like watched Star Trek when I was growing up, right? So, and so I've always had, you know, maybe you'll be, this will appeal to you, Ramon, considering your background. So Jean-Luc Picard, who is, you know, the captain of the Enterprise. And, you know, when you think about all the technology in on the Star Trek Enterprise, it's got all this stuff that, you know, it's got replicators and holograms, and you have, obviously an AI kind of visual, you know, voice agent, you know, and an ability to sort of move really quickly.

[00:11:25] And so, if you think about all the things that are possible, that we can imagine, I mean, human, you know, just the human mind is so, is amazing, that once you imagine something, it's inevitable that we're gonna go build it at some point. So the question is, you know, there's going to be an infinite number of things to go build, in my opinion, as we continuously progress. And, you know, Elon Musk talks about various levels of, you know, the scale of civilizations, et cetera. So as we keep moving in that direction, this literally like, you know, this is not some, okay, we're done in 10 years and we can relax. I mean, there's so much more to go build.

[00:11:57] And as this technology is used, I think it opens up new opportunities to go after. Not only, hey, like, do you have to deal with all the new systems now, or you got to think about LLMs, you got to think about, you know, great data foundations, et cetera, but there's so many new applications and possibilities that get unlocked. As an example, you know, if you think about drug discovery, you know, considering again, coming from kind of a family of doctors, it'd be amazing that, you know, if we can use AI to actually solve or cure some of the world's biggest ailments that debilitate a lot of people.

[00:12:36] So I think that's one of, you know, literally a countless number of examples that will get opened up here. And maybe one of the point I'd mention, Romain, is that the, you know, with the kind of the Cambrian explosion of companies that are gonna get started, and have been, if you look at any market map these days, they're like literally thousands of companies getting funded by venture capitalists, and including you folks, you know, you have your own startup fund. And so when you do that, all these companies are going after, you know, every possible part of the stack.

[00:13:00] I mean, whether that's chips, or whether that's LLMs, or whether that is, you know, applications, and that's, or compute, even if, you know, that's a foundational layer. So you literally have innovation on every level, because you're trying to optimize for this new possibility that you've now unleashed. So that just only creates more company, and it creates more demand for software developers, not less. Maybe certain large companies may have to optimize certain size of number of people to do what, but, you know, this is gonna be like, I think a lot of demand for software developers.

[00:13:28] Yeah, ambition just keeps increasing, right? When you have access to these tools, when you have like a bunch of AI teammates that are effectively like taking on like a reliable piece of work, you can do so much more. And then you just keep on dreaming bigger, going back to your dreams.

[00:13:44] And on the topic of like humans collaborating more as developers with like AI agents in some ways, like what would be your advice for developers to train to be good like AI managers of sorts? You know, like, I mean, I know, for instance, on my side and my team, we have completely changed the way we work this year. Like, you know, we rarely leave our desk without sending a task to like an AI agent because that would be a waste of time, you know, like, and so I'm curious, like, how do you think about this collaboration of developers managing agents in some ways?

[00:14:19] Yeah, I think, you know, it's actually, it's quite similar to how you'd manage teams. I think that when you're thinking about like the skills you need, you know, there's kind of a group of soft skills, and then I think there's a group of like critical thinking skills. And the soft skills are, how do you make sure that you are managing these kind of, you know, groups of, you know, now humans and agents, and humans using agents, and in a way that you can actually get to the outcome in a productive fashion?

[00:14:45] How is there a lot of cross collaboration between these various things? How is knowledge contextually rooted and shared? You know, we spend a lot of time with our own company, with SAP Oflows, internal product within companies.

[00:14:54] with Stack Overflow's internal product within companies, talking about the subject, how do they all work off the same context? And so, and then, and the kind of the more, you know, kind of like the, the critical thinking skills, I think is very, very important, because again, you want to learn or ask the right questions and prompt the right, you know, potential outcomes. And so the questions I think are equally important. I used the example of tests earlier on, like, you know, with testing, like how do you actually construct, even if it's an automated test that's being run by an AI agent, you know, how do you make sure that the agent is actually testing for the right things and not doing something super tactical, you know, and you're actually solving for the outcome that you bring? That takes a skill to be able to ask the right question or prompt the right sort of behavior of that agent to go do. So I think, yeah, critical thinking, and then soft skills. I think that's that combo.

[00:15:41] That's right. And I'm curious, like, whether it's like within Stack Overflow, or maybe like in one of your partners or customers, have you seen already, like, examples where, like, the human expertise combined with, like, AI kind of managed to solve something that you think humans alone would not have done?

[00:15:57] Yeah, I mean, we see this all the time, Raman, with our product, which is Stack Overflow Internal, which is the private version of Stack Overflow that thousands of companies use. All the big banks, all the big tech companies, retailers and all that. We have got the ability, and that AI functionality, by the way, of the product is powered by OpenAI, so we're proud to partner with folks on that basis. And what we notice is that our internal products, Stack Overflow, is used by all these companies. The APIs of that product were very, very hot over the past couple of years, and we went and asked what they were doing. They were actually using it for plugging into things like OpenAI's API for search, for surfacing content in various places, and building AI assistance.

[00:16:41] As an example, Ubergenie, Uber, the company, built an AI assistant called Ubergenie, which plugs into Stack Overflow's internal repository, the Stack Overflow, Stack Internal, as we call it in all these companies, and this assistant then goes into Slack channels and automatically answers a whole bunch of questions. And in certain other customers, they're going to support ticket channels, automatically answering a whole bunch of support tickets and closing them out. So that's a combination of the human context that is the institutional knowledge inside a company like Uber being leveraged by an AI assistant, soon an agent I'm sure, that does even more things with authority, and going and doing something pretty magical to reduce a lot of the noise in the system that's inside the company.

[00:17:26] That's amazing. And I mean, maybe zooming in on that, you mentioned, Uber, you mentioned partnering with banks also use your technology internally. From your vantage point, like Prashanth, like what are you hearing actually from these enterprise CTOs, CIOs, when they are trying to deploy AI for the engineering force, like, what are their challenges when they are trying to implement AI internally?

[00:18:02] Yeah, I think there are, first of all, I think there's this amazing momentum inside companies. I think, you know, maybe over 80% of, we have our own research and surveys and so on, like you folks, and we, you know, over 80% of them are actually doing it in some sort of meaningful way, whether that's experimentation or piloting. And there's obviously a smaller number that are scaling rapidly, and you know, folks who have completely gone to a stable place. And I think that spectrum, we see a ton of our customers in very squarely in the piloting phase, and, you know, even beyond, and in the scaling attempts, let's say.

[00:18:41] And I think the biggest thing that they sort of bring up is, you know, things that I'm sure you folks or your audience is aware of. So things like, you know, security concerns, you know, privacy concerns, governance concerns. It's the age old aspect of, you know, financial services companies or organizations are highly regulated. So for them to be able to navigate, to be able to really port an off access, to be able to really say that this is in fact, data is not being shared with another bank. You know, all the things that you, or other customers, all the things that you need to follow.

[00:19:22] And there's like a very long list of things. You know, I think those processes and approvals and all those things have not necessarily changed. Those all sort of stayed the same for a long time. It's new technology is pushing the limits of saying, look, like that's actually, some of these things are actually not valid anymore because the workflow is different. So, we see a lot of the standard security, privacy, governance kind of questions. And then the other thing we hear about is obviously the context. You know, they're very concerned about what data is being used inside the company to do whatever on the surface, for wherever the user is to perform an action, or even like, you know, do something even more robust to an agent.

[00:19:50] And I think they are finding it, they don't really trust that the data is accurate for those agents to go and kind of go and do what they need to do. So that I think is, I think, a bigger, and we obviously spend a lot of time in that area around knowledge management with our stack internal product. And so with our kind of human curation that we have in our platform, the combination of that along with the AI.

[00:19:52] Along with the AI functionality, including OpenAI's enterprise capabilities, that combination is a great combination for big enterprises to use, and that's why we see a lot of usage of our product in the enterprise.

[00:20:04] Yeah. That's awesome, yeah. One of my favorite pieces of, you know, data to get the pulse on the developer and engineer community each year is your Stack Overflow survey. Like your developer survey is always like an amazing way to see what's happening really in the world. Like not just on our kind of Twitter X bubble, but also like broadly in the industry.

[00:20:27] I'm curious, in that survey data, we still see some gaps between the trust and the usage of AI. And I'm curious, like from your vantage point, if you kind of have an instinct as to why do you think that gap still exists?

[00:20:40] Yeah, for sure. I think the stat that you're referencing is a very important one. We noted this this past year. So we've been asking the same AI questions for a few years and the two statistics that stood out to us this year were the usage of AI or the attempts or the interest to use AI have gone up even further. So it was only about 70 some percent a couple years ago, it's now in the high 80s. And so pretty much everybody in our community is using AI, trying it, interested in using it and so on.

[00:21:09] The trust level of AI actually was about 40% past couple of years and actually dropped about 29% this year. So about 10 percentage points, which was quite surprising. And when we actually understood what was going on, I think with the increasing attempts to use the tools, I think what has happened is that people, AI does hallucinate, it's not perfect as we all know. And it's got plenty of limitations.

[00:21:31] And I think with the particularly discerning population that we have, which are software developers, I think not, for example, not being used to the fact that what you're actually generating, you could have two different sort of pieces of code for the same prompt as an example, is like quite a little bit, it's quite jarring for people, especially when you're so deterministic and saying, I'm gonna do this, I'm gonna pop this out, and that's a satisfying experience when you're writing code manually.

[00:21:56] In this case, it's a little bit of it, it's a completely different way of writing code. So I think the trust level is one of not being used to it, probably number one. And in this notion of like it's, because it's firstly, it's not being used to it, it's not 100% accurate all the time. So you have to put in your thought, and is for a highly, let's say, sort of critical thinking person, I think they're gonna be like, hey, it's not perfect 100% of the time, so I'm not gonna push it into production, right?

[00:22:22] So those, I think, are probably, and maybe one other one I'd mention is that of course, there is the hype around job loss, which I think is never easy to say, am I using these AI tools to put myself out of a job, if that's effectively the other, I'm sure, overhang around why people don't trust it as much.

[00:22:44] Right, I'm also excited to see what the next survey will show, because I think from our vantage point, we've also seen that coding with AI has changed quite drastically in just the past two, three months, right? We talked a lot about 2025 as the year of agents, but AI agents that can reliably take on work, check their own work, edit on their work, write some tests, and all of that are becoming more reliable, so it's gonna be interesting to see how that changes that usage, but also, more fundamentally, that trust into what AI can accomplish.

[00:23:17] And if you have leaders in companies that have hundreds of engineers, or potentially thousands of engineers, how do you think these leaders should create an environment where AI outputs are increasingly more trusted, but also have some part of human accountability for all of this engineering force that is actually driving these AI agents to write software?

[00:23:40] Yeah, I think it needs to be, that's why we are a big proponent of attribution, which is, I think it's important to say, whatever the context that's being leveraged, is it leveraging off of human context inside these companies, which is obviously being produced by people, and making sure it's very clear that you can't, especially when you're pushing code into production, you can't just say, the AI tool actually made the mistake, so that's a problem.

[00:24:04] I mean, ultimately, these are tools, like everything else, and so, it's like, if you, back in the day, remember when we used calculators, you know, if you fat-fingered something on a calculator, and you got your exam question wrong on you, it's not on the calculator. So I think similarly here, like you have access to this tool, so use it, and just know that you have to own the outcome.

[00:24:23] So you better learn how to use it well, know when something goes wrong, that you can catch it when it goes wrong, in the hallucination situation, and write good tests, and just make sure that you're thorough. I totally agree, exciting times, for sure.

[00:24:38] Switching gears maybe a little bit to the culture of learning for developers, like Stack Overflow has been known for its tough love, and I was curious to like see how you're rethinking that culture in the age of AI.

[00:24:50] that culture in the age of AI and maybe more like an empathy-driven design. [00:24:56] Yeah, no, great, great point. Tough love, school of hard knocks, you know, and we appreciate those descriptors. [00:25:03] And, you know, the brilliant part of how the founders set it up, you know, back in 2008 of the company was that the idea was to make sure that the quality was extremely high, and that was the intent behind it, and came at a cost which is to say it's not gonna be like this super soft place where you're just gonna be able to have a conversation. [00:25:21] It's not that, it is like this objective of getting, you know, the accurate knowledge and all things software programming. That was the original intent. [00:25:30] So what we have done more recently to your point is that the world has obviously changed. So we have done several things, including, for example, on the public site, we have really incorporated AI to assist the users to actually ask the question. [00:25:44] So, for example, when you ask the question on Stack Overview, when you asked the question previously, when I asked my first question, I got slapped on the wrist and said, look, it's a duplicate question. We're gonna close it out. [00:25:54] In this case, it is gonna tell you, the AI is gonna tell you on the public site, hey, look, it's already been asked. So you may want to sort of word it differently. Or what about these? Go check these out. [00:26:06] Those answer your questions. Even before the user is going to be in the town square with all the humans in the School of Hard Knocks to get that tough love, the AI is gonna be giving you a little bit friendly kind of coaching behind the scenes. [00:26:20] So we've done that. We've done something called staging ground where people can actually get a place where they can get their questions refined and actually volunteers go in there and help them with their questions. [00:26:29] This is what the human to human sort of experience. We've opened up several new other non Q&A areas, Raman, which is new for us. [00:26:36] So, Stack Overflow is known, as you're aware, for this canonical Q&A sort of format. Perfect question, perfect answer. Which by the way, has been amazing for things like LLM pre-training, as you folks are partners of ours on that basis also. [00:26:51] And I know this is useful. Sure, thanks for being a great partner on that as well. Yes, no, pleasure doing it. [00:26:56] And so in that case of Q&A has been very, very useful because it's extremely high quality data. But what we have now recently opened is things like opinion based questions. [00:27:05] So people can actually have, there may be multiple answers to a question. You want to have a discussion of best practice, what are some various ways to actually debate it. [00:27:12] So there's a lot more of a discussion. So there's areas, it's the same area, and you can add these new types of tags to be able to sort of separate from the Q&A. [00:27:21] And then we also opened up human chat where you're able to have various, and in fact I just posted this on my LinkedIn today. [00:27:27] So we have one of our users who had an excellent experience who said, look, this is like revolutionary. I have a very collaborative experience with another expert on this topic. [00:27:35] And they were probably talking about OpenAI's API. And they're having a very real time discussion. It's a live chat where you're talking to another expert in the room. [00:27:46] And they're helping each other out as a fellow community member. And that's what's magical about our platform is that the intent is to help others. [00:27:53] It's that you've had to sort of broaden the form factor. Yeah, it's really great how you've expanded to bring more of that learning, bring more of that nuance, too, where it's not just like one answer being approved and the other necessarily being incorrect. [00:28:06] That's right. And I have to say the last one, which I forgot to mention, is our biggest innovation around this topic is around what we call Stack Overflow AI Assist, which is also powered by OpenAI, by the way. [00:28:20] So if you go to stackoverflow.com and you click on the top left nav, you click on AI Assist, it is a very Jack GPT type life experience, which you folks have pioneered with this beautiful single prompt window. [00:28:34] And you're able to ask any question, what do you want to learn today? But it's going to be rooted in the 85 million questions and answers on Stack Overflow. [00:28:43] And it's going to surface the context, keep it in context as you're having the conversation. And it's going to be able to not only use a rag to be able to do that, but also fills in the gaps that we don't know on our site, but let's say OpenAI's model does in GPT-5. [00:28:57] It's able to produce all that in one recommendation. So that is, you know, that's a beautiful experience. A lot of great feedback from our users. [00:29:04] And we've seen that, you know, some really increasing usage, that's good. That's fantastic. [00:29:08] And I guess with all of these new like touchpoints that you're adding to those Stack Overflow experience, and most of them driven by AI, how do you think like AI can help developers feel maybe less judged when they're learning and maybe like having more inclined in like being more confident to learn in public with others? [00:29:26] Yeah, this is actually a big, you know, our mission statement is cultivate community, power learning, and unlock growth. [00:29:33] And, you know, it's all focused on the user on the public platform. This is by the way, for our public platform, we obviously have a completely separate set of objectives for our enterprise business, which is Stack internal. [00:29:42] But for the public platform to make sure that they are actually get, we are focused heavily on, you know, we've always had a reputation system.

[00:29:48] you know, we've always had a reputation system and people who accrue reputation points. A lot of people get jobs off those, you know, their profiles, they're able to proudly showcase them in other places.

[00:29:57] And so we want to make sure that their contributions in also these new form factors, they're accruing, you know, good reputation points, ultimately they are constantly on that journey of learning, including the newer technologies. So including how do you really prompt, you know, CodeX is an example to like, you know, do a really good job, you know, with whatever you're writing.

[00:30:18] As long as you're helping other people, I think that accrues to you. So I think it's just, we're focused on making sure the user is getting connected to their fellow community members in new ways, that they are constantly learning the new way of doing things in software programming. And that should ultimately accrue to them in terms of unlocking that period growth.

[00:30:35] Right, well, when we started the conversation, you also mentioned something, which I think is very critical to remember, which is like, you have to master these AI tools as developers, but you also can't quite forget about the fundamentals of, like, how the machinery works behind the scenes.

[00:30:53] What advice would you give to universities on the other side when they're kind of like, rethinking their curriculum for their students, their future engineers, in this kind of new set of AI tools and AI Xeroid industry?

[00:31:09] Yeah, no, it's a little bit of a fortuitous question. I sit on both the board of visitors of my undergrad, as well as our board of advisors of our Cornell Engineering School. And we recently, they asked, they brought us together to answer this question. They said, hey, like, what are your suggestions on how should we incorporate AI into the curriculum?

[00:31:34] And I think the answer, at least mine, was around the workflow. It's like, when you think about what we're actually doing, when you're teaching students, first of all, university education, I think, needs to be re-hauled to make sure it's actually useful and practical as people graduate.

[00:31:44] It's always just been disconnected. Like, what you learn in college is completely out of date by the time you come out of university, typically. So in this case, it's like, hey, assume that these AI tools are part of your workflow, and if that's the case, then what would you actually do to actually go from A to B in whatever engineering topic, you know, as one example?

[00:32:07] So I think rethinking the workflows and the pedagogy in a way that says, you're always having this kind of assistant. It's almost like, you always have, now you have a calculator and math exams. So you always have this when you're doing your work. How would you actually accelerate? How would you utilize the insights and personalization, the ability to get summarized insights from a lot of different data sources, all those things, and then perform actions against that?

[00:32:31] And so a lot of it was around redefining the workflow and the context of AI. It was sort of my advice to them in the curriculum.

[00:32:38] Yep. That makes a ton of sense. And if you were to put yourself in the shoes of a mentor for an engineer, what would you tell them to focus on first?

[00:32:47] You know, I go back to the same advice of like, you know, make sure you don't skip any steps. It's like, you know, because if you do that, then you're gonna be absolutely, I think, you know, you're gonna be invaluable to any organization because, you know, you've got different classes of people that are coming in.

[00:33:02] People who are the veteran software developers who are excellent and, you know, have learned it the hard way, et cetera, and are kind of in the system. And then you have this new batch of folks. And there's a batch of folks who are likely gonna skip all the fundamentals and just, you know, bytecode their way into like the next part of their career.

[00:33:20] But then they're going to probably hit a wall at some point when they discover that they may not have actually learned, you know, much. They've only like learned how to like bytecode it really well. So I think the combo of both these things is that third cohort, if you can actually, you know, do both, like learn the way, a new way of doing things with AI tools, but also learn the fundamentals, then I think you're going to be gold, you know, in later on in your career.

[00:33:42] So, I think that's great advice. Maybe to close this off, like looking ahead to the future, like what excites you the most about like, you know, the steep curve of these model capabilities for coding, how like you get like teammates now that can like work alongside you and this kind of like collaboration between human and AI coders.

[00:34:02] Like what's the most top of mind, as the most exciting thing for you?

[00:34:07] Yeah, I think there are two things. I think we are very excited about our own business, obviously at Stack Overflow, with our Stack Internal product, which is now all about this knowledge intelligence layer that's now being plugged into every company where they're able to leverage it as sort of the base of context that technologies organizations to then leverage for their agents to do all sorts of actions.

[00:34:26] So that's great, that's very exciting. And, you know, great to partner with you folks on that dimension. The other part of course is I think the broader implications of what you're talking about, which is that, you know, just the rate of change and the ability to unlock these new possibilities, especially in like high impact areas, whether that's in healthcare, whether that is.

[00:34:46] whether that's in healthcare, whether that is in life-defining sort of areas. The applications that are gonna get built, I think are gonna be absolutely spectacular because you're gonna have so much more, the barrier to innovation has dropped even further. I mean, already it was pretty low, but this is like a completely new level because of all the compounding S-curves that are sitting on each other. And I think that is just gonna be spectacular to watch the innovations that come out of, in real hard impact areas like biology, like drug discovery, I'm most excited about that. I think especially as we think about longevity and so on.

[00:35:19] Amazing, well, I think that's a great note to close. Thank you so, so much, Prashanth, for sharing your insights with us. Thank you so much for sharing more of your vantage points from the Stack Overflow side and we are very grateful for the partnership between OpenAI and Stack Overflow and we can't wait to see what we can keep on building together.

[00:35:40] Same Romain, thank you so much for your questions and for your partnership and for OpenAI's partnership with Stack Overflow. We're really, really happy with it, so thank you. Very much likewise.

[00:35:49] Thank you so much, Prashanth, and we'll see you soon. See you soon.