The State of the Web

Hi, everybody. Welcome to JS Nation. My name is Sasha Grief, and today I'm going to talk about the state of the web. And now, I realize this is a pretty broad topic. So more specifically, what I have learned about the state of the web through running web development surveys. So what do I mean by that? Every year, I organize a bunch of open surveys online. Anybody can take them. The biggest one is probably the state of JS, but there's also state of CSS, state of HTML.

And just out of curiosity, raise your hand if you've taken one of these surveys before. Okay, so quite a few people. So first of all, thank you for participating, because if you didn't, I probably wouldn't be here talking about it today. But I don't have that much time, so I want to jump right in and talk about the state of JS. The latest edition took place in November of last year. And as you might imagine, I asked developers a lot of questions. And that's actually what I want to do with you guys today as well.

This is not going to be a boring old talk. It's going to be a live interactive quiz. Now, I know what you're thinking. They just showed you another quiz. I didn't know that. I thought I was being very original and very innovative. But at least my quiz won't have any trick questions about Greta Thunberg or Billie Eilish. And so, well, you might be thinking, how do I participate? How do I, you know, keep track of my score? And you can do that using something almost all of us have today, which is this, your fingers. So you can count on your fingers. We will all start with three lives, and then for every wrong answer, you lose one. And we'll see who makes it all the way to the end. There will only be ten questions. So hopefully a bunch of you make it through. But let's start with a really easy one. What were the three most used front end frameworks in the latest state of JS survey? So A, Svelte, Solid, HTMLX.

All great libraries, but maybe not the most widely used. B, Backbone, Knockout, and Mootles. Now, if you've never heard of these, don't worry. You haven't missed the newest front end library coming out the past couple weeks. It just means you're probably under 30 years old. And then C, React, Vue, Angular. So I think it's easy enough to figure this one out. Of course, it's C, React, Vue, Angular.

So in the survey, we asked not only what respondents used, but also how they feel about it. People could say if they like or dislike the framework they used. And one framework in particular had a lot of positive opinions and came in first. So which one do you think it might be? Give you a little, a couple seconds to figure it out. The answer was C, Solid. So if we look at the data, these are the State of JS results. Solid is actually that yellow orange line all the way at the top. Not only was it number one last year, it has been number one for the last four editions. Number two is Svelte, the green line. Number three is Vue.js making a comeback to get number three. These are rankings. But if you look at the absolute value, you can actually see that most libraries are towards the top of their charts here above 70% satisfaction.

I think that's interesting, because we often have this image of JavaScript developers complaining about their framework and being quite negative. But the truth is when you look at the data, people actually enjoy for the most part using those frameworks. I think that's a testament to the work being done every day by the maintainers of each of these projects. So let's talk about TypeScript. Among survey respondents, there are more developers who A, only use JavaScript, B, use 50% JS, 50% TypeScript, or C, only use TypeScript. This is among survey respondents, right? It's not necessarily among the entirety of the JavaScript population throughout the world. But among the subset of respondents who took the survey, turns out the largest group is C, people who only use TypeScript.

Among survey respondents, there are more developers who only use JavaScript, use 50% JS, 50% TypeScript, or only use TypeScript. This is among survey respondents, not necessarily among the entirety of the JavaScript population. The largest group among survey respondents is people who only use TypeScript. The chart shows a significant number of people using 100% TypeScript, indicating its mainstream adoption and becoming the default for web development.

The TC39 committee decides on JavaScript language features, considering various proposals annually. One of the proposals listed is not real: explicit resource management, temporal, reverse threading, or decorators. The fake proposal is reverse threading. The top three upcoming proposals respondents are excited about are temporal, a new way to handle dates and times, and decorators. Temporal, the most anticipated, aims to reduce reliance on third-party libraries like Moment.js.

A survey question asked about the favorite new CSS feature among respondents, with 'has' being the most enjoyed. 'Has' is a CSS feature targeting elements based on their children, offering new possibilities in CSS. The CSS 'has' selector is supported by all major browsers as of June 2025. The baseline indicator confirms its support in all four browsers since December 2023, making it safe to use. Surveys like the state of CSS provide detailed information on features, baseline status, and additional resources for developers.

So okay, don't run away. I know it's not CSS nation today, but I still wanted to include a couple questions about CSS, and in the survey, we asked respondents which feature they had enjoyed the most over the past year. So not necessarily something that came out in the past year, just something that they started using. And so which one do you think was number one? Has, add container, subgrid, or float write? It's probably not going to be float write, because even back then, nobody enjoyed it that much. But the real answer is actually A, has. Now has is a really cool new CSS feature that lets you basically target an element, not based on its parents, like you would usually do in CSS, but on its children, so what it contains, and it really opens the door to doing a lot of new things with CSS. And so that was people's favorite new feature.

But you might be thinking, well, that's all well and good to have all these new fancy CSS features, but it's not really that useful if you can't actually use them in browsers. So that's what this next question is about. As of June 2025, which of the four major browsers support the CSS has selector? A, none of them. B, only Chrome. C, Chrome and Firefox, or finally, D, all four. So which browser support has? And the answer is actually D, all four. So there's actually this thing called baseline indicator, which is a standard convention developed by browser vendors to let us know what's safe to use or not, and has the status newly available, meaning it is supported in all four browsers, and has been in fact since December 2023.

By the way, when you are browsing through the state of CSS survey results, you can actually click any feature and get extra info on it, including the baseline status and also links to MDN, can I use, and other resources. So another survey we did last year was the state of HTML, and I want to stay on the topic of baseline. So here's another question. If the baseline indicator says a feature is widely available, which is the final stage of baseline, it means it's A, supported by a majority of users, B, used by over 50% of the top 100 websites, C, supported by a major browser for 30 months, or D, available at your local supermarket. So probably not D.

But the real answer is actually going to be C, supported by major browsers for 30 months. And that's the case of a feature like aspect ratio, another CSS property, for example, which might seem like it's fairly new, but in fact it's been over two years, 30 months, since it became newly available, since it became well-supported. So this issue of browser interoperability is something that we ask a lot about in the surveys, including here this question in the state of HTML survey, which is about the features that people would want to use but cannot, because of poor browser interoperability, which is really hard to say, or browser support. And number one here is the popover API followed by anchor positioning and view transition API. Now what's interesting to know, I think, is that there's actually an initiative from browser vendors called interop, which is that every year they get together, so the Chrome team, Query, Edge, Firefox, and so on, and they decide which areas to focus on to fix those problems so that we can all use these cool new features. And so in 2025, the focus areas are anchor positioning, backdrop filter, color web vitals.

And what's cool is that one of the ways that they prioritize these features, one of the data points they consider when, you know, drafting that roadmap is the data from the state of HTML survey, the state of CSS survey, and so on. So I think it's quite cool to know that the browser vendors are actually receptive to feedback and that one of the ways you can make your voice heard is by completing these surveys. But let's switch gears now and talk a little bit about AI. So we had another survey called the state of web dev AI, and there's a ton of AI surveys out there. A lot of them are even called the state of AI, so I'm not the only one. But what makes this one different is it's not a survey of AI developers, it's a survey of how web developers, like all of us, basically, use AI or maybe don't use AI, because that's really what I wanted to figure out. Like how common is AI use really?

So here's a question for you guys. What percentage of respondents to the survey do you think generate the majority of their code via AI? So not even, like, all their code, just more than half of their code. I have to say, the result kind of surprised me when I saw it, because it turns out only 22% of respondents fit in that bucket. So you can see here at the top of the chart is people who do not use AI at all, and at the bottom is the people who use AI for everything, and there's only 11 of them. The majority is in the middle or actually towards the top of the chart, and that's my case speaking personally. I use chat GPT from time to time to maybe generate a helper function or a component or some styles, but I haven't used, you know, cursor or any of the newer tools to, you know, vibe code entire apps just before breakfast, like you see on social media.

And I think that's still the majority of people in that case. So if you are like me, like a lot of survey respondents, still coding apps, so, you know, by hand, don't feel bad about it, don't feel any, you know, FOMO towards the AI crowd because I think we are still the most common use case.

And finally, the last survey I want to talk about today is the state of devs. So this one took place in April of 2025. So it's very fresh and in fact it's so fresh that the data isn't even out yet, so you're getting an exclusive sneak preview of those results. The other thing that makes this survey special is that unlike the other surveys which were about JS, CSS, AI, this one was not about code at all. Instead, it was about all the other things that make us who we are, you know, your career, your workplace, but even things like your health and hobbies.

So as part of that survey, one thing we asked about was your job title. And it turns out there's one kind of job title in particular which is correlated with a higher income. So do you think it's A, developer? B, engineer? C, webmaster? Or D, consultant? Now, as much as I personally would like webmaster to make a comeback, I think it's, you know, the 90s are cool again these days, so who knows. Probably not going to be C. But the answer it turns out is B, engineer.

So if you look at the data, here are job titles with median income. And if you look at all the engineer jobs, they tend to be at the top while the developer jobs tend to be more towards the bottom. So with the caveat that correlation is not causation, just to be safe, I would suggest updating your resume and switching from frontend developer to frontend engineer if you can.

And finally, our last question of the day. The top hobby enjoyed by developers is A, video games, movies, reading, or sport. Now, hey, don't laugh. That's not fair. A lot of developers do enjoy sport. In fact, I do bouldering, so if anybody want to go climbing later, let's go together. But no, sadly, the stereotype is true. So for extra credit, does anybody want to guess what the most popular specific video game was? Let's just yell out some video game names. Pac-Man? No, sorry, maybe in the 1973 survey, Pac-Man would have won, but in the 2025 survey, the result was Baldur's Gate 3. That game came out not that recently, but I think it's so long that people are still playing it. Number two was Valatro, which I also played, but I had to stop pretty quickly because I could feel myself getting addicted by the minute.

So let's count how many lives remaining we have. If you have zero lives remaining, you didn't make it to the end, sorry, but you can still call yourself a developer. Now, if you have one life remaining, you survived, congratulations. You can upgrade your title to architect. So instead of front-end developer, you are now a front-end architect. It's the same thing, same job, but it sounds a lot fancier. If you have two lives remaining, you only made a single mistake, so you are a very reliable person, you can be lead, front-end lead, and lead us to our promised JavaScript future. We'll follow you. But now, if you have all your lives remaining, raise your hand, let's see. Nobody? Really? I'm sure there's going to be at least one person. So hypothetically, if somebody had survived with all their lives remaining, they would deserve the ultimate title of engineer.

So what have we learned today? Probably not much, but still. You can use more new CSS features than you think. Thanks to tools like Can I Use and Baseline, it's easy to keep track of new CSS features and the browser vendors are doing a really good job these days of making sure that they quickly became supported. TypeScript is now the default. It took me a really long time to start using TypeScript myself, but I think it's just a way the ecosystem is going at this point. Most developers don't use AI that much, at least not yet. People are talking about developers being replaced by AI. I don't think that's going to happen, because after all, there will still need to be people who write code manually, if only for AI to be able to plagiarize it.

And then finally, the wave is changing fast and surveys can help you keep up, not only because when you take the survey, you are exposed to a lot of new features and libraries that you might not know about, but also when you review the results, you can learn about upcoming trends and the results themselves can serve as a jumping off point to browse other resources. Speaking of taking surveys, there is a survey open right now, the state of CSS 2025, so you can go ahead and take it right now, or maybe not right now, wait until the end of the event. But yeah, I wanted to thank you for taking the surveys. Thank you for participating in my little quiz, and I hope you have a great JS Nation. Thanks.

Yeah, let's kick it off with just a question I have for you. What is the biggest challenge in reaching people with this survey? Wow, that's a great question. So I think it's a dual challenge. One challenge is just reaching as many people as I can, right? And so you can rely on things like podcasts, I don't know if you know any good podcasts about web development, I might spread the word. But then another challenge is making sure you don't reach an audience that is too biased in one direction. For example, if let's say Foo.js posts a link to the survey on their homepage, then Foo.js is going to be overrepresented in the survey. And that's even more true for things like gender diversity, country diversity, racial diversity. So if you want to increase those factors and make sure that the survey represents a broad population, then actually being more deliberate about how you reach out to people is also important. So I'm still looking for the right answer to that question, but it's definitely a big challenge every year.

So you've been doing these surveys a long time now. In your mind, what's the most surprising thing overall that has shown up in any of the data? The fact that I'm still doing them? Yeah. In the data, I think the positivity overall, people don't hate JavaScript as much as you'd think. I don't know. Maybe things have changed, because I think a lot of really great work has been put into the ecosystem, but eight, nine years ago, we were all talking about JavaScript fatigue, right? And the idea that picking a framework is too hard, there's too many new frameworks, and so on. But if you look at the data over all these past years, first of all, there aren't that many new frameworks. React, Vue, have been around for a decade. Svelte has been around for five years, or something. So things are not changing that much, at least in those terms. And then people actually really appreciate those tools. So I would say that was a really nice surprise.

Yeah. All right. It looks like I can actually get on the Wi-Fi now. Also, folks, if you're changing tracks, could we make sure we keep it down a little bit, so that way we can all hear Sasha? So let's see. Let's do this first one. Is the low AI usage a we are not allowed to use AI? What do you think? Yeah, good question. That's not the impression I have. But I'm also not sure that's something that the survey specifically asked about. So I could definitely ask about that next year. That's a really good idea. But I think the impression I get is that people just haven't started using AI more than they are not allowed to. The survey also asked about not only do you use AI, yes, no, but which tools you ask. And so chatGBD, for example, is definitely used a lot, as you might expect. But things like cursor or co-generators like Bolt or P0, those are used far less. So to me, it seems more like chatGBD is becoming mainstream, but there's still some lag for more specialized tools. And I would expect that lag to diminish so the other tools to become more popular as well. But I still think it will happen at a slower rate than what AI proponents might be predicting.

Let's see, we got another one here. How many participants do you have per survey, and which of the surveys is the most successful? So I think the most successful ever was the situp.js a couple years back when we had over 20,000 participants.

These days, it really depends because if one popular YouTuber like Fireship or whoever mentions the survey, that can drive a couple thousand respondents. So it's very, very unstable, I would say. But usually around 10,000 respondents for the biggest surveys and maybe 5,000 for the smaller ones.

Do you need to do a lot of manual work to normalize the survey data? Yeah, I'm sure there's a lot of responses that are all kind of similar, right, in text form? Yeah, I feel whoever asked that question is like sitting behind me when I work at home because that's a big part of what I do. So the survey can have pretty fine questions where you pick an answer from checkboxes, radio buttons, and so on. But it also has a lot of free form questions where you just type in your answer. And so there's a lot of work required to transform that into usable data. But I am trying to automate that as much as possible.

How do you select questions subjectively? So what happens is for every survey, I will put up a preview of the survey. So actually going back even further, at the end of the survey, you can suggest new questions for the next edition. So there's a specific question asking people how they would like to see the survey evolve. There's also going to be a GitHub issue asking people to suggest new ideas. And that issue is open, like, the whole... I opened the issue for 2026 in 2025, basically, so they have one year to give feedback. Based on those two things, I will then create a survey preview for the upcoming edition of a survey, and that will stay open for about a month.

How do you select questions subjectively? So what happens is for every survey, I will put up a preview of the survey. So actually going back even further, at the end of the survey, you can suggest new questions for the next edition. So there's a specific question asking people how they would like to see the survey evolve. There's also going to be a GitHub issue asking people to suggest new ideas. And that issue is open, like, the whole... I opened the issue for 2026 in 2025, basically, so they have one year to give feedback. Based on those two things, I will then create a survey preview for the upcoming edition of a survey, and that will stay open for about a month.

And during that time period, people can suggest new questions, comment on existing questions. And at the end of the preview period, then I finalize the survey and launch it. So people have three separate chances to affect what the next survey will look like and make their opinion known. And so it's a mix of, of course, my own subjectivity, I guess. Because I'm still curating the questions, but also community input. Yeah, and you've had a lot of experience doing these. Yes, I guess so.

So next question here is, I thought 22% using AI over half of their code is actually huge. Is there a selection bias towards enthusiasts? So that's a great question. It's something I try to avoid. So having, you know, of course, if you do a survey about AI, people who like AI are more likely to do it. Because people who hate AI will say, hey, it's not for me. And in our emails to our mailing list, I specifically said, please take the survey, even if you hate AI. If you don't use AI, we also want to hear your voice. So hopefully it mitigated the bias a little. But of course, you can't really mitigate it completely.

Nice. Next one here. Why did you start doing these surveys? So that's something people asked me a lot. And back then, I was using a framework called Meteor JS. I don't know if you've ever heard about it. Big fan. And Meteor JS, the whole point was that it was all in one. So it did the server, client, you know, templating, whatever.

The problem is when I stopped using it, when I ventured out of that bubble, I was very confused by the broader JavaScript ecosystem. And I know you can relate. And so that was the original impulse for the first survey. I wanted to figure out for myself which technology to invest my time in after being, you know, spending so much time in a very self-contained ecosystem.

Nice. Well, thank you so much, Sasha. This has been amazing. Thank you. Great talk.