The 2025 Stack Overflow Developer Survey with Jody Bailey and Erin Yepis

0:00The Stack Overflow Developer Survey is an annual survey conducted by Stack Overflow that gathers comprehensive insights from developers around the world. It offers a valuable snapshot of the global developer community, covering a wide range of topics such as preferred programming languages, tools, and technologies. Jodi Bailey is the Chief Product and Technology Officer at Stack Overflow, and Aaron Yepes is a Research Manager at Stack Overflow. They join the show with Sean Falconer to talk about the results of the 2025 Developer Survey, which was recently released. This episode is hosted by Sean Falconer.

0:39Check the show notes for more information on Sean's work and where to find him.

0:55Aaron and Jody, welcome to the show. Thank you. Great to be here, Sean. Yeah, and Aaron, I guess I should say welcome back. This is now our annual podcasting event where we discuss the Stack Overflow Developer Survey. So I'm excited about that. Good. Yeah, let's go ahead and get this set up for 2026. Yeah. Get it on the boats early. Exactly. So I think we got a lot to cover. And before we get into it, I wanted you to help sort of frame what the Stack Overflow Developer Survey is for those that are not familiar with it. So can you give a quick explanation of what it is and what you're trying to accomplish with that?

1:32Yeah, so the Stack Overflow Developer Survey has existed for almost as long as Stack Overflow, the site, has existed. So this has been our 15th year of running it. It started as a very small thing. There's a lot of people on the site asking questions just to get that really clear, aggregated view of what technologies are you using? We can see what you're asking about, but it was just the most obvious question to ask that community. And since then, it's grown into quite a bigger deal than that. But it's just a unique opportunity every year to get that temperature check from actual developers, from them directly saying themselves, instead of a third party inferring what it is that developers use and care about and do.

2:20And it means a lot to our community. What's the typical number of people are participating in sort of the makeup of those developer profiles? So for the most part, the core group is one of the questions that we ask is, do you code for work? So a majority of our users every year indicate that they are a developer by profession. So they're working as developers somewhere this year. It was 76 % indicated that. But the survey is open to people that are students as well, or people that are adjacent in a way, like their PMs, project managers, and stuff like that. Besides that, we have a huge audience, obviously, here in the U.S.

3:00of all the countries that responded this year, last year, and three years ago. The U.S., Germany, India, U.K. are always in the top five. This year, about 50 % of all respondents came from those four countries in addition to France and Canada. So we have a strong European, North American, and Indian developer group. And as we saw this year, what we have been seeing trending lately too, is that we have a lot of more experienced developers. So we ask this every year, how many years have you been coding? 2025, 65 % of the respondents indicated they've been coding for more than 10 years. And Jody, did you have something you wanted to add about your perspective on why you do this and the value that it brings to the community?

3:46Yeah, so I bring a slightly different perspective because I was one of the people that was always looking at the dev survey as a dev manager, as a developer, even understanding what other people are doing. What are the trends? What can I learn? And as a hiring manager, it was always important to understand what tools and technologies people were using. So it was always a barometer. But it was something I always looked forward to kind of see what was new and exciting, what people were thinking, and was super helpful just in terms of identifying trends, as well as thinking about how I hire and create a place that people wanted to be a part of.

4:24Yeah. And I was saying, you know, sort of in the pre-show that I reference the survey all the time. I think that if you really want to start sound smart in conversations, there's a lot of meat on the bone that you can leverage there for all kinds of different questions people might ask. Like, how's Rust doing as a language? It's right there in the survey that you can sound smart quoting some statistics around that. So that's one of the things like a lot of the places I've worked, we've been polyglot where there's lots of different tools and technologies. Right. And everybody wants to use something new and cool, Rust.

4:55Right. And the survey would be a good reference. OK, well, if we're going to consider this, what is our hiring goal? Right. I mean, do we really want to have a tool that maybe we can't find enough developers to work in? and end up with this code base after our Rust, not to pick on Rust, but fill in the blank language after that person, that advocate leaves, then what do we do? So the developer survey was super helpful in that way too. Yeah, yeah. Is this like just a moment in time trend that we should invest in this company and how does that impact us? Or is this something that's going to be, everybody's moving this way.

5:29I think the survey results certainly can help with those kinds of decisions within the company. From your perspective, and I'd love to hear from both of you, if you kind of have different views on this, like what is the biggest story coming out of this year's developer survey? And maybe we'll go back to you, Erin. Yeah, so I think the biggest story, and I would not be alone in thinking this, is the numbers we have around adoption, trust, and favorability. So for the third year, we see more developers than ever are using AI tools. And I should have mentioned that. Trust of AI tools. Not just of AI tools.

6:05I knew what you were talking about. Yeah, because I referenced this so much, I'm like, you guys know what I'm talking about. Because it was a pretty big story. So we see that for the third year in a row, we have this growing adoption of AI tools. Now we're over 80 % of our respondents, 84%, I believe, are using AI tools for work. But the trust took a dramatic dive this year compared to 2023 and 2024. And favorability did as well. So it's interesting to see that, wow, so more people are using these tools, but they don't like them and they trust them less. To me, I think it makes sense as you have, one, there's just more AI tools being developed.

6:48So you have a lot of things that you can try, a lot of things that could fit into your nuances of what your role is as a developer. So not just a one size fits all type of deal. But I think we have these high expectations as well for certain developers. And of course, everyone knows when you have high expectations, there's more probability that you're going to be disappointed. I will say this, that when we look at, specifically, we asked this year to use AI tools daily, weekly, less than weekly, like what's the variability in use there? There is a correlation that we see for the developers that indicated they use them daily.

7:25They have higher favorability scores than those that use AI tools less than daily. So I think it probably speaks to more of that learning curve. Like, hey, I just tried a couple of things. Some of it didn't meet my expectations. But if I keep up with it and find out how to make this work, then I can figure out my expectations are at the right place. I'm using the tool for the right thing that works for me. And we also see with younger developers, developers that have less career experience, they have higher favorability and trust scores as well, which probably also just speaks to, hey, my bar was low already.

7:59I don't know what I don't know because I just started doing this job. So there's room to grow there as far as like those two things. And as I mentioned, we have a lot of very experienced developers that answer the developer survey. Of course, their expectations are going to be high. Yeah. So the key is set low expectations and then you'll be blown away. It's kind of like when you go see a movie and everybody told you it was amazing. And you're like, it's okay. Yeah. Yeah. Nothing disappoints you like high expectations. So I would think too, as part of this, we're becoming more comfortable with AI.

8:33It feels probably less magical because people are using it day to day. So you're less blown away by it. And then also, I think if you're using it all the time, you naturally encounter the rough edges more often. The other thing I wonder, which could be a factor in terms of the decrease in trust and sort of people enjoying these tools is that maybe in the early days, essentially the collection of developers that were using them were more the early adopters that were excited about this. And now a lot of companies are sort of taking the stance that, hey, if you're going to develop in our company, you have to use this tool.

9:06We purchased it. We want the efficiency gains. And now you're kind of, whether you want it or not, you're kind of being forced to use it, which naturally also creates sort of a negative reaction to it initially. Yeah, I think that's spot on. It's something I've seen, not just survey wise, but just talking to peers and other people in the industry. And when people, as you mentioned, are forced to use it, so to speak, especially a tool like this, where my observation and experience, it's not like you're perfect right out of the gate, right? I mean, you have to learn how to use the AI tools and how to break things down into manageable pieces.

9:44And people are still experimenting as they're building things. I talk to developers and it's like, well, I've got to figure out ahead of time, is it going to be more work debugging the code I generate or just easier to write it myself? And it's figuring out what that balance is. And going back to the high expectations and the dictation, nobody likes to be told what to do. or how to do their job, and especially developers, right? I mean, we've always considered ourselves craftspeople where it's as much art as science. And then to have somebody tell you, oh, well, just generate a bunch of code and it'll be fine, that it discounts the value of what developers bring to their job.

10:27And there's also these super high expectations that all of a sudden we're going to have this massive increase in throughput. And if you think about it, and I always encountered this with pair programming or mob programming for my team, it's like there's this assumption, especially by leadership or people that aren't in the business, that developers spend eight hours a day typing code, right? And we all know that's not the case, right? So if you think about the fact that developers spend 20 % to 40 % of their time writing code, Even if you improve the throughput there by cutting it in half, it's still a small percentage increase in the throughput.

11:07That said, I think that's where there's real opportunity in terms of how to leverage AI. And what I'm seeing with some of the thought leaders within my own team is identifying ways to reduce time on those things that aren't about writing code. The things that, whether it's collapsing meetings or if it's about figuring out how to think about designs and break things down more effectively, you see oftentimes those types of things using AI are really beneficial. So what I'm trying to do and encourage with my team is experiment with the different tools, find the ways that work best for you and help you, and not necessarily, hey, all code has to be written with an agent.

11:49Yeah, no, that's a really good point about the fact that an engineer's time is not hands-on keyboard writing code all the time. So you're only going to get so much efficiency gain there, even if the tool is as successful as possible. I think one area too that I'm seeing a lot of investment from companies, I can't remember if there was anything covered in the survey, but I'm curious to hear your thoughts on is the bane of many teams' existence is on-call duty. And there's a lot of companies and investment on the AI front there of how do we alleviate some of the stress of being on-call, provide better tooling to solution those.

12:25And a lot of times too, if you're an on-call engineer, you might be on-call for systems that you're not necessarily working in all the time. So you end up inevitably having to go to someone who's like an expert in the company to ask them questions. And then you're sort of taking them away from their time that they can be thinking about designs or executing whatever their job is. So it ends up being this sort of cascading effect of sucking up a lot of people's times, not just the person on call, it's all these other people that might be involved in it. Yeah, so we didn't have a specific question about like that use case, which I do think is interesting, like kind of, yeah, that support aspect of things.

13:00But I will say that we did ask about what are like specific tasks, what are you using AI tools for? And we created this visualization where it's sort of superimposed the tasks that people said they're mostly using AI tools for and the satisfaction that they have with AI tools. And we can see that for the most part, the people that use AI tools a lot are using it for searching for answers, which I think would be when you can't find the experts or the expert is not available. That's the use case. And if it's helping them, they are more satisfied with that use case, too, of all of the things that you can be doing with AI tools right now.

13:40So not necessarily related to the survey, but just my own experience and working with the team is using AI to explain code and break things down has been really useful. So you can imagine in that use case, and I see my quality engineers doing this to reverse engineer code or as we're looking to break down really complex pieces of code, using AI in that way is really helpful. And you can kind of picture that as well from a support perspective, because you're right. It's like you have one person and you can't expect them to know the whole system. So how can you help break it down, make it easier to understand?

14:15I think there's real opportunity there. We see our infrastructure teams focused on improving those aspects. And that applies to new people on the team as well, or even junior developers too, just helping to sort of grasp where things are, how the code's structured, having sort of a psychologically safe zone to ask. and AI automation might be in favor to having to ask somebody who's on the team. In terms of AI agents, there's a tremendous amount of hype right now around AI agents. And there's a lot of stats out there that indicate this could really be the next frontier software, but the survey shows that adoption is still relatively low.

14:54What do you think is holding developers back there? So yeah, we had a couple of questions where we asked about, in your own words, And for AI agents, we had this too. We asked about like, what is the impact for those that have adopted it? So sort of like backwards engineering that what we see for those that have adopted it is that they were able to see productivity gates. and they were able to learn new things faster and upskill, reskill. So I think where the adoption is lacking is that whoever is in charge of either purchasing or making the decision, they're not able to connect the, this is a possibility.

15:39What does it mean for in time and money? And what's the ROI on that? So there's people that have already taken the leap and they're able to show like, hey, my experience as a developer, this has helped me do this and this. It has saved me time. It has helped me learn. And it's just like that, I guess, because we see generally surface level, a lot of feedback about AI, it's hard to connect that to the specific agents. Should I invest in that? Is that worth the time of standing it up, debugging a couple of things with integrations, get it running on a couple of local machines, having the team all in on it.

16:19We're all like on the same page. Like, yeah, it's going to take some time to invest, but we're seeing more of the good parts. And we see that in the dev survey too, is that they've said, I use this for the people that have used it. Yeah. I have seen these impacts. Yeah. One of the challenges, and it goes back to the hype, right? Is the hype is way out in front of the actual engineering and the learning curve. And it's easy to forget, or maybe not, how fast things are changing, right? And how, especially as engineers, we're still just learning how everything works. I mean, MCP as a protocol is really new, but it's really powerful and it's changing the way people think.

16:58Building agents is... It's only a year old. Right? But it feels old already, right? It reminds me of, and you probably remember when JavaScript was taking off and it felt like there was a new framework every week, right? It's like you didn't know what to do or how to do it. And it's kind of the same thing. So you've got the hype and you've got all the people out on social media talking about, oh, it's doing this. I did this in five minutes and you can do all that. But meanwhile, all the engineers are trying to deliver all the things they are already committed to, trying to learn new tools, what it can really do versus what people say it can do.

17:31And then how do you scale it, et cetera. So I'm seeing already people are starting to adopt and build agents that didn't know what an agent was six months ago. Right. And so I think two years ago, no one knew what a vector was. And then suddenly, yeah, it was a line with two points, right? You know, from math and physics, right? But. Tired of babysitting autoscalers and overspending on cloud costs? Meet Thoris, the platform that makes engineers heroes to their finance and business leaders. Thoris intelligently manages Kubernetes clusters, automatically right-sizing and scaling workloads while preventing downtime from traffic spikes.

18:11It anticipates usage and capacity needs so systems stay fast, reliable, and efficient without constant tuning. Teams using Thoris cut cloud spend by 40 to 60%. Thoris predicts compute and GPU demand before it happens, keeping performance smooth and costs in check. Stop wasting compute and guessing your resource needs. Let Thoris handle your auto-scaling so your teams can focus on building. Find out how much you can save with Thoris. Visit thoris.ai and try our cloud savings calculator. So another trend that I think the survey talks about that also is something that we see a lot on social media and so forth besides AI agents is this concept of VibeCoding.

18:54So first of all, for those unfamiliar, can you explain what VibeCoding is and why do you feel it's kind of resonating with some of the new developers that are out there? Yeah, so VibeCoding is just a terminology and it's cauterized in Wikipedia. I feel like it has some sentiment associated with it. I wanted to make sure whenever I asked the question, I'm referring to this thing. This is a social phenomenon. I'm not trying to put any shade on your work or what you do for whatever, however you get your job done. So it comes with a lot of, I feel like, very passionate emotions for some developers.

19:33But vibe coding is just you're using these AI tools to sort of tell you what to do when you may have like zero to little knowledge about what it is that you're trying to accomplish. There's a lot of people that don't even have to touch a keyboard in order to create a whole app. They can just speech to text and then it does everything else in the background. So there's like levels to it. But at its core, it's just you're using these tools where your knowledge is not quite there. You're using it to bridge the gap in order to get something accomplished software code wise. Mm-hmm. Yeah, I think the survey said 77 % of developers say vibe coding isn't really part of professional work.

20:14So do you think that this is primarily about enabling a new class of user to create things or focus on certain types of side projects versus something that we'll see in a professional setting? Yeah, definitely. I think that for right now, though, the phrase itself isn't helping. It sort of just conjures up an immediate reaction. And we saw that. but in their own words, because this question on the survey was an open response. So I let them say whatever they wanted to. And then I used an NLP algorithm to sort of group it all together. So I have like responses that were like emphatic. No, I am not vibe coding.

20:55Yeah. Did you get any sentiment analysis on that? I'm sure there's wild swings. There's lots of good stuff there. But I think so another indication and Jody already mentioned this earlier too, vibe coding is associated, I think, with this future tech debt. So your co-workers, your peers don't want to hear about you pushing this vibe coding idea because to them, they're like, that's just something I have to fix later. And we know that, yeah, 45 % of developers indicated that debugging AI generated code was their top frustration with using AI tools. So outside of AI, based on the survey, are there any technologies, frameworks, or shifts that kind of stood out in the data that people are excited about?

21:40Some of the things that I've noticed as far as what's been growing in usage and popularity for technologies, I mean, it's not new, but we see, I created this new section. So we kind of grouped all of the cloud DevOps tools, technologies, platforms together. And we see that now that we know when people can focus on it, because a lot of these tools, especially when you look at some of the things that are being used for AI as well, it's old stuff that's being repurposed. It's like, oh, yeah, this is perfect for this use case, too. We are seeing this just like growing usage of cloud technology. We saw a huge jump in people using Docker and Kubernetes, which again, these have been around for a while.

22:24They've been on the survey for a while, but I just think we're now getting to like that point where everyone, it's just becoming more standardized. Postgres SQL, you know, again, an old favorite, but it's coming back the spotlight because of AI. That's just being kind of repackaged or remarketed in a way for that. So we saw a huge jump in usage for that this year, too. There was also in the last year, really, that sentiment around like Postgres for everything movement. We don't need all these other databases, just use Postgres to solve all these problems. So I wonder if that kind of impacted the responses that we saw in the survey.

22:57Yeah, probably. Again, because we have a majority of more experienced developers. They've been around. They've seen like, hey, if you made it this far in my career, probably this is a good thing to keep using. So instead of moving to something that's just brand new. But yeah, other than that, it was also interesting to see. So Stack Overflow itself, we have a new section where we asked about which of these new popular tags, what are you interested in as far as the technology goes? So while people associate Stack Overflow with the top three HTML questions, JavaScript questions, Python questions, we see a growing interest and a lot of popularity, a lot of people that are using these technologies for those AI tools.

23:43So large language models as a tag is growing in usage. MCP growing in usage. We also see Rack as a tag growing in usage. So that was pretty cool. It shows, again, it's that temperature check on the developer community that we have access to. Was there surprises in the survey? You know, things that came up that are different than prior years that you weren't expecting? Yes. So this year, I added a question at the very end. So if you made it all the way to the end, which a lot of people did, I just asked general survey feedback. Like, hey, was this too long for you? and then two, open response, general feedback.

24:27And I got a lot of feedback. So it was pretty cool to see that they took the time. Again, they took the whole survey and then were able to leave feedback. But a lot of that feedback was general annoyance of having to answer so many questions about AI. Yeah, I think that's kind of some of the things that we touched on earlier. I think there is like a, it's natural where there's such a saturation of something in a market for people to start to become negative towards. It's like overexposure of even in media is like, okay, we have no game shows and then we have too many game shows or we have no reality TV and then we have too much reality TV and then people become turned off.

25:04And we're probably going through the downturn in, I don't know, comic book movies right now. It's like, okay, well, we got oversaturated. Now we need to contract the market essentially. And maybe we're seeing some of that with AI as well. What do you think engineering leaders should take away from some of the findings around team productivity, training, tool adoption? How should people be thinking about that? I think similar to things I've been saying that in terms of adoption, engineering leaders need to think about how do they get their engineering teams to leverage AI to create the outcomes that they want, which isn't necessarily to write more code faster.