Stack Overflow CEO Prashanth Chandrasekar sat down with Ryan at HumanX 2025 to talk about how Stack is integrating AI into its public platform, the enormous importance of a high-quality knowledge base in your AI journey, how AI tools are empowering junior developers to build better software, and much more.
Read more about Stack Overflow’s future here.
Learn more about HumanX here. Missed it this year? The event takes place again on April 7-9, 2026 in San Francisco. Early birds can register here.
Follow Prashanth on LinkedIn or explore his posts on our blog.
[intro music plays]
Ryan Donovan Hello everyone, and welcome to the Stack Overflow Podcast, a place to talk all things software and technology. I'm here at the HumanX conference with Stack Overflow CEO, Prashanth Chandrasekar, and we are going to be talking a little bit about the show, about AI in general, and how Stack Overflow fits into it. So welcome to the show.
Prashanth Chandrasekar Thank you, Ryan. Wonderful to be here with you, and the HumanX conference has been a phenomenal event for us, just realizing how much of an impact that we are making and also how much we can really lean into the movement of agentic AI and help developers and companies. So this is a really exciting time for us.
RD Yeah, I think it's interesting. I don't think a lot of people look at us and think ‘AI,’ but there's a lot of ways that AI can touch our site and our products.
PC Yeah. I mean, literally in all parts of our business now, obviously we're trying to be very thoughtful about how we play in the AI wave, and obviously to make sure that we are really focused on solving problems and we're not sort of jumping on the hype train for the sake of it, but what value can we add as the world moves forward to take advantage of this amazing potential technology capability? And so if you just take it in multiple parts, the public platform which is Stack Overflow, which has been around since 2008, over 60 million questions, thousands of questions being asked every week, and we have millions of people going on there every month, and that is really, really powerful because the quality of the data that you've collected through a lot of painstaking curation is phenomenally useful in the context of leveraging and for pre-training LLM models. Now even increasingly inside companies people want access to this knowledge base and data for their AI agent experiments that they're running and pilots, so that's been very interesting to see companies reach out to us for that. And then especially on that platform, what we are doing is trying to be a lot more of a headless Stack Overflow. People remember us as going to Google and landing on a Stack Overflow link. Now, a lot more people go to AI chatbots like ChatGPT. So what we are doing is, obviously through all our partnerships we are surfacing Stack Overflow content in all these places that people are spending time in, our technology and developers. So we want to go wherever the developers and technologists are, so if they're in ChatGPT versus Google, no problem, we will show up in ChatGPT. We will also show up in Google index search results because that will continue to be a source, but any other AI chatbot and so on. The other thing that I think we have realized is that we want to open up the number of lanes on the highway, so to speak, on the public platform because the users that we talk to in places like this, they all have varying needs. They love the Q&A format and we should never pollute that because it's very important for that to be independent of any sort of AI slop, etc, and so we will never do that, but there's an openness on experimenting with new content types, things like chat and maybe even discussions and maybe even other forms of media on top of Stack Overflow in a kind of unique, distinct parts of Stack Overflow, so that developers of all kinds. I met with a high schooler yesterday who was so excited about still leveraging our platform but he has very specific mediums through which he would like to engage on our platform as an example. So the transformation of Stack Overflow on the public platform already is very useful because the quality of the Q&A. We will maintain that, we will add other sources over time, other content types. They may be of lower age, lower quality, but higher frequency, and even maybe more ephemeral and that I think is very powerful. So that's all in the public platform, and then our enterprise products with Stack Overflow for Teams, which is a private knowledge base for companies to leverage, is a really huge boon for advanced companies building AI agents inside their companies or AI assistants, depending on your definition of these words, and they're basically tapping into Stack Overflow's APIs, Stack Overflow for Teams’ APIs, along with the public platforms APIs, to really build powerful agents inside their companies so that these are working off really high quality data that is stored in these private repositories on Stack Overflow for Teams inside their companies, and that's phenomenal. And we have companies like Uber who has written about this publicly, which is really, really powerful for us to see, and we're encouraged that a lot of customers are moving in this direction and leveraging this sort of phenomenally really accurate knowledge base. The knowledge solutions business is also all our AI partnerships with OpenAI and Google and the other hyperscalers, and you'll see us make a bunch of announcements here. We're working with several more and several of those partnerships are coming along nicely, which require them to license our data as we've already talked about and surface that content into their AI tools. And even by the way, empowering those developers to ask questions over time into Stack Overflow from those spaces, but at a minimum, get links to Stack Overflow when they're cited and attributed. And of course, our final business is advertising, and what is very clear when you come to shows like this is all the AI startups, et cetera, that have been funded by venture capital, et cetera, they all want distribution. They want to be able to get their products in front of developers, and what better place to do that than Stack Overflow, because that's where the developers spend time. So even that business I think is a net beneficiary of AI. So literally public platform being transformed into this headless Stack Overflow, the private knowledge base of Stack Overflow for Teams is a great source of raw material for agentic needs, and then your knowledge solutions is all our AI partnerships, then you have advertising business, which is focused on now these AI companies looking to promote their products. So literally all parts of our business are tapping into this new world of AI and agentic AI and so on.
RD Yeah, I think it's really interesting. Obviously, a lot of these LLMs were training on our data as we released the public datasets all the time, to go to them and ask for attribution and make sure that, I think that's really valuable to ensure the community is vibrant.
PC Yes, absolutely. It's on more than one way, Ryan, because if you notice, we both talked about this statistic that there's a lot of enthusiasm in our community to use AI tools. I think 70 to 75 percent of the folks that we have polled over the past couple of years have said that they would like to use AI tools or are interested or very enthusiastic about using them, but only 40 percent trust what's coming out of these AI tools. So for them, the missing link is, “Okay, how do I know I'm not going to get fired if I use the answer from this AI bot inside my company?” And so the hype is catching up. And so in that context, when people are put to test and put your neck on the line, so to speak, by using these AI tools– because you can't blame the AI tool, it's on you. Whatever tool you use is up to you, and AI is a tool. And so in this case, if they use it, they want to make sure they can rely on it. Hopefully they know the fundamentals, so they haven't skipped them so they can understand when something is obviously wrong. But in this case, by providing attribution in links and citations, you're grounding these AI answers in real truth. You can rely on the truth on Stack Overflow, because again, thanks to our users who spent countless hours curating knowledge and continue to do so, you can trust that that's been well-answered. So that's, I think, really the power of attribution. That's why we really insisted, also it's a licensing requirement that we have with our partners to say, “You have to attribute back to Stack Overflow links,” that way we can be proud to be associated with that because it's not hallucinating and we can stand behind whatever the answer is in that chat interaction.
RD That trust gap is an interesting one. I think when we first started talking about it on the podcast, everyone was talking about RAG because that gives you sources, that shows the work of the LLM. But I know you've seen it too that there is a fair amount of pushback from people in the community about AI. How do you plan to sort of assuage those fears of the AI slop, the AI takeover, whatever?
PC Yeah, I think the main thing to do there is to make sure that we don't pollute high quality knowledge bases. We're literally one of the few places that has high quality knowledge bases, so our commitment is never to do that, right? We never want to pollute our Q&A platform. That lane is super high quality. Yes, it takes time to curate it. It's not instantaneous and it's not instant gratification, but the reality is that there's a great outcome from that patience. But at the same time, as I mentioned previously, by opening up those other lanes, whether that's chat or discussions or anything else, it gives people the opportunity to have maybe ephemeral conversations in the moment, serendipitous conversations with experts on the platform that just happen to be in the room where you're discussing, because that's where you want to just sort of riff on the latest technology development about the latest DeepSeek model that came out. And so we want to enable that sort of thing to happen on Stack so that it is expanding beyond, but we want to draw very clear lines of demarcation between what is the canonical Q&A base, which is pristine, et cetera, and then you've got everything else, which is not to say it's not good, because you still have accurates. All right, so you have experts in those places. However, it's not going to be perfectly worded, these are ephemeral conversations, et cetera, and there may be even opinions, right? They may not be the canonical answers, but it may be okay for maybe users just want directional guidance from humans when they're stuck on something very quickly. This empowers that.
RD Some of the best learnings I've ever had about software development is going on something like Hacker News and reading comments and hearing them fight it out. And I know we've added some AI tools that help people make better questions. There's the question assistant that we've released to the general public. What do you think about the AI assist stuff that we're adding?
PC Yes. So I think in the context of thoughtfully incorporating AI into our public platform, what we're very happy to announce is that now all questions will have Google Gemini, in a friendly manner, nudging question askers on the best way to write the question, suggestions, “Hey, this has actually been asked already,” because it's going to be trained off of the public platform data, so it’s just to serve the user in a way that is not exposing them to what has historically been a very consistent feedback item from our community, saying, “Hey, I'm a new user and I just want to ask a question. I just got slapped on the wrist because some expert on the platform said this is a duplicate question or it wasn't well worded,” et cetera, because there's a reason why we've set up those rules because we want the quality to be super high, so it is a school of hard knocks on that dimension. But the great benefit of leveraging something like Google Gemini in this case is to give users a private sort of feedback loop in a friendly fashion before getting sort of publicly sort of exposed on asking let's say a non-ideal question. Now, by the way, that was my experience when I first started on Stack Overflow, so I'm very pleased that we are now doing this because it's a super thoughtful implementation, so we're happy about it.
[music plays]
RD Developers and infrastructure engineers drowning in noisy logs? Mezmo’s free trial helps you filter, enrich, and route log data in real time, cutting noise and improving efficiency. Spend less time searching, more time solving. Try Mezmo free at mezmo.com/freetrial.
[music plays]
RD It's also using some traditional ML techniques in addition to the Google Gemini which is an interesting sort of thing. I worked on the blog post about it. In terms of integrating with the sort of real time data stuff, how do you think that's going to be used? We talked about inference time data a lot, the reasoning models.
PC Yeah, I think this is exactly the reason I think a lot of folks, if we look at who's actually reaching out to us for access to our APIs and data, both on our public platform as well as all our enterprise customers, beyond the big AI labs and the cloud providers who obviously we are constructing partnerships with pretty much all of them, the large enterprises who are experimenting with building agents inside their companies want the latest and greatest for things like to make sure their agents are up to speed, to leverage concepts like RAG and indexing so that they can actually leverage the latest information when they are providing a recommendation to a person that has a question inside the company, and that's just powerful because we are one of the few places where that real time accurate data is being produced on the internet as well as inside Stack Overflow for Teams inside their companies, if they are customers of ours, because that's where real time information is being shared inside companies, which, by the way, plugs into your Slack, plugs into Stack, Microsoft Teams, and so on, as you know, and so it is tapping in accurate, canonical information or answers to questions inside companies and that creates a great engine for building these agents.
RD And we have some interesting partnerships and you're going to be on stage today talking about some of them. Can you give us a little preview or post-view as this will be out in a few weeks?
PC Yeah, we're very excited to say that we have now integrated with Microsoft Graph Connector. Microsoft's one of our great partners and we love working with them. They've been partners with us for many, many years. And this is on the heels of us announcing our GitHub Copilot integration as well where our content of Stack Overflow data shows up in GitHub Copilot as an agent, basically, which empowers developers on GitHub Copilot to get access to all this rich information within that platform. But the Graph Connector now plugs into the Microsoft ecosystem of Copilot tools, which is really, really powerful because anywhere Microsoft Copilot is used, Stack Overflow data can be leveraged both on Stack Overflow for Team's content, which is obviously very, very useful inside companies, as well as our public Stack Overflow data because we can plug our APIs into that framework. So that's a super exciting way to, again, realize this vision of being a headless version of Stack Overflow that surfaces up into all the tools that people spend time in.
RD I think that's really interesting, the agentic model. I've had a couple of people here talk about it almost in terms of the microservice moment for AI, and it's very cool that we're going to be plugging into that system.
PC Exactly right. And we've gone through so many waves around this, right? We can break up into tiny parts and then reassemble them into big parts, that's been going on for decades. So we just happen to be in that phase of the cycle, so we'll see how long it lasts this time around.
RD That's very exciting. So we’ve both been at this conference since Monday, Sunday. What has been some of the most exciting things you've seen here?
PC I think generally speaking, I was happy to see that there was a real dose of realism baked into the hype cycle of AI agents and the buzzwords of the year, so to speak. Not to say that those concepts are not visionary because they are, because imagine having agents that sit side by side with you and me helping us with our daily work. And that's actually pretty awesome that that gets realized. And I use AI tools, I'm sure you do. Before I get onto a customer meeting, I'm pretty quick to get to speed on the customer using things like that, and that's all very useful. It's efficiency, it's productivity, right? So I'm generally a fan. You can call me a techno-realist, maybe leaning into an optimist, but I was very happy to see that sort of balance in this conference. It is termed HumanX for a reason, and who are we serving? Here in our case, we're serving developers and technologists. That's what we've got to remember. And AI is a tool to empower these developers and technologists to do an amazing job beyond even what they're doing today to make their lives easier and to free up time for them to go solve more technical problems or something else. And so it's good to see that when people acknowledge that it's easier said than done. When people want to roll out ‘AI agents’ inside companies, it's not as simple. I mean, we're still going through the first wave of even thinking about how do you leverage Gen AI inside companies beyond the initial pilot group that has been experimenting with it for 18 months. And they're very happy, but what about the rest of the company in your large organization that has got all the baggage of not accurate data and process of cement and you've got people that are worried about whether or not they're going to have jobs after this? There's all sorts of issues that are all very standard issues. There are people issues, there are process issues, and the technology issues are actually the least of our concerns. I mean, it's actually all the other stuff. And so, yes, so much more needs to happen for this technology to be truly realized at a price point that can be accessible, but there's plenty more to go solve. So I was happy to hear that, both in my interactions and looking forward to talking about that more here.
RD Yeah, I talked to somebody in VC the other day and he said we're in the first inning, and I appreciate, like you said, the grounded, because somebody's got to implement those AI agents, somebody's got to debug them. There's a real code behind it. It's not just this airy pie in the sky stuff. Do you think there's going to be a rubber hits the road moment where people find out what the limits are for these AI agents?
PC Yeah, I think that people will realize that some parts of the stack have been sort of, people have gone ahead and some are going to be long poles in the tent, so to speak. There are actually constraints to be able to realize the full vision. So yes, I think this is when people will try to push hard on making it happen. I don't know any companies that are running production-grade agents that are running around the company for reviews. It's actually more of a vision than it is a reality today. But I suspect a lot of people will try to do early versions of it. An example is what I've mentioned previously– we have advanced companies in our customer base on Stack Overflow for Teams leveraging– I don't know if I would call them AI agents. They don't, they call them AI bots or AI assistants, and the word agent itself I think requires clarification what exactly we mean by that, because I think it's misused. But generally speaking, the idea is to empower developers with powerful assistance, at least for now, assistance as in just that, and empower them to do a great job with their daily work, right? And so that I think you'll see a lot more of this year. Whether or not they actually are given the full agency to complete tasks on your behalf–
RD Or whether they should.
PC Or whether they should, I think all that stuff, I don't think is a near term solve. I don't think anybody's anywhere close to saying, “Yes, let's hand over the keys.”
RD So what are you excited about the future of AI? What are you thrilled about?
PC Well, I think the potential of what this could accomplish I think is pretty amazing. I mean, if you think about even beyond just software development, I think the potential for it to discover drugs for diseases that have long plagued humanity, I think is an example of something that I think is absolutely fascinating. We all have people that were affected in our lives with chronic diseases and so the fact that you have this immense amount of ability to process information that already exists and build on existing information, do reasoning, even though it may hallucinate a lot today and make a lot of mistakes, but you can imagine over time that gets better. And even if it has a, I don't know, 50 percent accuracy rate, there's a better chance of solving problems in a much faster way than maybe waiting years to get up to a solution. So I'm very happy about that. I think it's a really promising part of what this is all about, and especially in some fields when they come together, could create this really compounding effect. We talked about transformer technology and we talked about chips really getting advanced and we talked about obviously algorithms, cloud compute being available everywhere. All those things are great and now even with something like quantum, which I don't necessarily spend a lot of time thinking about but I haven't read about it, it suggests that if something like that has immense fundamental sort of shifts in progress when combined with what we're talking about here in AI. So a couple of these things, when they come together at the right place, right time, I think creates the opportunity for just immense progress, which I think is just phenomenal. It does have its risks and downsides and everything else and the risk of getting overhyped and so on, but there is absolutely a thread there, a scenario there that you can imagine that could be very, very productive for society. So that's probably what I'm most excited about.
[music plays]
RD It's a hopeful note to end on, but we're at the end of the show. I have been Ryan Donovan. I edit the blog, host the podcast here at Stack Overflow. If you liked what you heard or didn't like what you heard, comments, suggestions, questions, topics, cover, email us at podcast@stackoverflow.com, and you can reach out to me on LinkedIn.
PC Hello, everyone. Thanks for listening. My name is Prashanth Chandrasekar. I'm the CEO of Stack Overflow. Thank you for being part of our community, and keep your requests and suggestions coming on how we can better serve you. You can reach out to me on LinkedIn with my name and also on X. Thank you.
RD All right. Thank you very much for listening, and we'll talk to you next time.
[outro music plays]