A Look Back at Web Dev in 2025

Welcome to Syntax. We've got two episodes for you coming up in the next week. Today, we have our reviewing our 2025 prediction. So every single year, what we do is we look at where Wes development is, and we say, these are our predictions for the next year. What's gonna happen? What's gonna flop? What's gonna pop? And we are gonna go through all of our predictions today and say, did we hit it or did we not? And then next Wednesday, we'll have another episode, which is our predictions for 2026 as to what's gonna happen. So my name is Wes Bos. Happy holidays. Merry Christmas, everybody. Hopefully, you're enjoying it. This is gonna be a really fun episode. I'm here with Scott Tolinski. How are you doing today, Scott? Hey. I'm doing good. Did you just come up with that did it flop or did it not on the fly just now? I I'm I'm just genius. Yeah. Yeah. I just have an incredible intuition and all of that. Oh, yeah. No. That's great. I, I like that a lot. Did it flop or did it not? That's a great, metric I think we can all,

by here. Or did it pop? Clop or pop? These episodes because this is the thing I like doing the most, just pontificating on, like, theoretical who knows what. Is this gonna happen or not? I I feel like I'm always a little too optimistic on stuff. So,

the reason why Last year, we had a lot of, like man, we are a little up too optimistic on that. But I I have to say that this year, we we nailed quite a bit. So let's get on into the first one, which was the temporal API will ship in the browser. So I said both Chrome and Safari will ship by end of year, and Scott said that it will be in at least one browser. So if you haven't heard of it, temporal is this massive API that is set to replace the date API. You might be saying, well, what's wrong with the date API? Stop changing stuff. And I don't know if anybody's ever said what's wrong with the date API. Yeah. That's I don't Node. That's Fair.

Fair. So the temporal API will allow you to encapsulate durations. So if you've ever wanted to, say, I I want to somehow make three days, how can you say that in a in a a standardized way? Or if I want to do January 25 to January 26, but regardless of time zones. Or if you do want to include time zones in a start and an end, there's just time is time is hard. Time is complex. What is time? And the temporal API is going to solve all of these. It's gonna be a absolutely killer API.

And where are we at right now? I think we nailed it. So here's where we're at. May 26, it was added to Firefox.

Good. December 3, it entered into Chrome beta, which means that a a month later, January 7, we're going to 01/07/2026, we're gonna be it's gonna be landing in Chrome Stable, and then it's currently in Safari behind a flag. So, like, we've been talking about temporal for, what, like, four Yarn? And I feel like we're

we're pretty much there. I'm super excited for it. I like Temporal, and I've been using it with the polyfill for, like, three years now. So, you know, I've been using it for a while, and, I I I've shipped stuff with that polyfill. So, yeah, I I'm stoked about it, but I I this is the type of movement that it feels like I've been wanting to see for a long time, and and here we're at. Yeah.

You can start to learn temporal. You know? Our next one is that on device AI becomes common. On device small AI models, OCR detection becomes much more ubiquitous and developer friendly, running either in the browser or on user devices. What we did see in browser AI stuff in Chrome, I feel like a lot of these types of features, they kind of, they go under the radar a little bit. We are seeing more OCR stuff. We are seeing more, speech to text, text to speech.

We are seeing all those things.

As far as, like, small model AI, I think that's happening more often than we realize.

I don't think it's as common. I'm gonna say we miss this, and this is someone I love, transformers JS. I've been playing with a lot of the built in Chrome stuff where you can use a local model in Chrome. I don't think we're there quite yet. I think people care much more about it's not that expensive to hit an API.

Right. But I I think as people start to say, like, what will what does it mean when you don't have to pay anything for it, when it just runs on your user's device? What could you you could run it constantly in a loop. You know? So I I don't think we're quite there just yet. You are seeing more a little bit more niche models. You Node? Sanity detection is a really good one where Node you need to detect if something is spam or something's Mhmm. Someone's saying something really off. Chrome is trying to figure out what the API looks like. What does it look like to have structured inputs and outputs? They're really just playing with a lot of that right now, same way transformers JS. So I'm gonna pump that to Wes year. You say this. I'm a say well, like, but and and I'm saying that to someone that's used them several times this year. But if we're going by what the prediction actually was, it becomes

more common. More common than what? It was before. Yeah. That's a win to me. Yeah. You're right. It's not it's not it's not that it is I'm trying to I'm trying to get us some points here, Wes. I think it became more common.

Whether or not that became ubiquitous. I guess we did say much more ubiquitous, not it will become ubiquitous. Yep. So I think we saved ourself with the wording here.

I was talking to some of the, engineers that work on Chrome, and we're talking about this. Like like, what are the benefits to running on device? Like, there's obviously speed, there's privacy, and there's cost. Right? And right now, I I don't think that people necessarily care a ton about cost. It's way faster to use a hosted model than it is to use a local Node, plus they're way better. And, unfortunately, I don't think a lot of people necessarily care that much about the privacy aspect, which kinda pains me to say, but I think that's just the reality.

Yeah. Well, the next one is the Wes GPU unlocks fast local machine learning.

And I do not I think it might be a little bit early band, but type GPU is doing some cool stuff where they even have, like, an example of text prediction and stuff using WebGPU.

I I don't think

enough people are doing this just yet. People are doing it, but I don't think enough people are doing it. The WebGPU stuff just dropped in all the browsers, and WebGPU is what's gonna make it fast in the in browser. Like, I have an example where I generated alt text for, with transformers. Wes. And the Wasm example versus the Wes GPU example is six times faster.

And, like, then it moves from a cool demo to, like, oh, we could actually use this. So I think that's the one that's the one that we're gonna hopefully see more of in the next year. But that's that's the next show. I'm not gonna talk about that.

We're not making any predictions right now. Yes. So let's talk about AI. So our

big prediction was that the the models will plateau in quality, and we're gonna see big improvements on how the products use the existing models.

And I think we nailed that, but, like, let's let's dive into, like, the different aspects of, like, where are we with AI versus three hundred and sixty five days ago. Right? Because in some regards, the models didn't get that much better. Right? Like, we saw incremental improvements. You think they did? Okay. So I think they did.

So tell me about this. I think if we're comparing

how accurate the responses of a raw input versus output of something like, what is the latest one from from Anthropic? Yeah. It, it is a Sonnet seven? Or Opus 4.5 or SONNET 4.5.

I think OPUS 4.5 is raw input versus raw output is light years better and more reliable than what we had last year.

I I just do. I trust the output of Opus 4.5 so much, endlessly more, with all kinds of tasks.

It's still not great at CSS. It's still not great at some things, using browser standards, but the end result, I think, is much better. Okay. Input output.

The input output. So I think that the actual tools that we use with these things got significantly better. So the agents the, like, the quality of code that agents are able to to generate, even with if you're comparing them with the models that we had at the end of last year, the output of them are so much better because of the tooling built around it. You know? Like, the cursor, Copilot, all of these agentic tools that sort of go back and forth and reason with themselves, the quality of that has gotten significantly better. In in fact, like, I Wes, like, halfway through this year thinking about this prediction that I had, and I was like, oh, boy.

Was I wrong about that? Because I was like, I I think we'll see incremental improvements, but, like, I feel like the actual quality of the code that these things output, like you said, is significantly better. And and I I'm attributing that just to tools, but maybe I'm wrong about that. It's it's more model based.

I I think it's both. I think there's a a big increase in both because we did see a tremendous improvement in the tools themselves.

But I do think if we're if I'm going to, you know Yeah. Like I said, Opus 4.5 and I'm giving it the same question, I think I would like the answer enough that it makes me feel like it's a huge leap because we're, like, last year, I I wouldn't trust it at all. Last year, it was, like, kind of a fun experiment.

I would make fun of vibe coding. This year, I can vibe code personal software using these tools and actually feel like it's not total garbage. In fact, I I've gone back with projects that I would have vibe coded last year using the latest model, and it could clean up the code substantially better. And that is part tools, but I also think the models is a big part of the Node. JS well. Now let's talk about stuff that's not coding. Like, video

and photo gen has gotten unbelievably good. Right? Like, we saw Unfortunately, shock. Unfortunately, Sora, nano banana, it's a weird place because it's they're unbelievably good, yet a very not that useful.

There obviously are some use cases for it. Like, we're designing a a jacket right now, and I've been going back and forth doing some mock ups with that and giving getting ideas, and I find that to be super useful. But, like, aside, like, Saura was funny for, like, a week and a half and and now it's just like and a half. Like, boomers getting tricked on Facebook by videos that have the watermark removed and, like, nano bananas, people just posting, like like, fake stuff. Like, I think it's like an at least right now, this stuff is a massive net negative in terms of, like, just trust. And is there any good use for these, like, lifelike videos aside from it being absolutely hilarious?

I don't even think it's funny. Like I said, it was entertaining for, like, an hour and a half, and I was like, oh, that's funny or it's creepy or it's weird. Yeah. But, like, man, I can't imagine more than that being like, I made this now.

It's like, oh, it was really funny when I made Wes Boss in The Sopranos looking like Pikachu. Okay. Okay. That's that's very funny to me. But there But Sora is freaking so funny, though. Like, the one where the the

pig comes through the window and the man's sitting on a toilet.

Oh, man. I still laugh about that, like, monthly. There's something that I'm going to like meme funny. You know? Like, people are just, like, constantly creating new memes and pushing this thing to the limit, but, like, not all that useful.

Yeah. Especially, like, use energy and stuff that you use to make me laugh. Yeah? It's useful in creating really bad commercials.

Man, I when I see, like, an AI commercial or something Yeah. The problem is is that the people creating a lot of this stuff, the same reason why it sucks is the same reason why the people creating it were not creatives or commercial creators before. They're like the business people who are all like, oh, now I can make a commercial with AI, but they have no artistic vision at all. So that whatever they put into the AI is still gonna come out sloppy, because they don't have an artistic vision for it. Yeah. You don't have craftspeople

using this stuff. It's the marketers or the grifters doing it. That was actually one thing about Sora JS that Sora somehow figured out how to make the videos funny. But then, like like, after a while, you kinda you kinda understood all the hilariousness, you know, and it only knows how to you you find all the, like, tells, you know, like, quick talking back and forth, like a cutoff at the end of something. Those are what makes it funny. But then, I don't know, after a couple weeks, you realize, oh, that's just the

the recipe for making it funny. Yeah. Next one. Text to UI tools get really good. Things like bolt dot new, v zero, etcetera, will get much better and faster.

They get so good that some people stop using Figma at least for early prototypes.

Yeah.

I don't use them still because I think that's, like, the part of the project that I personally want to have a hand in the most is the early stage parts of it. But I do feel like the tools got, I wouldn't say really good. They got better. Yeah.

I think my bias talking.

I think we we nailed this prediction because Cursor, Versus Code, antigravity, Chrome's MCP, Puppeteer, all of these agents now have the ability to take the UI into account. Both, like, screenshotting it, both taking videos JS well as, like, clicking on things. And and and that whole loop, I feel like, makes the debugging step and, like, prototyping step a lot better. I don't know that, like like, tools, like, explicitly lovable bolt and v Deno.

I haven't seen that grow at all in the, like, dev space, but I I know it has grown quite a bit in the, like, Node space, the tinkerer space. You know? And that's a that's a whole new aspect of people who are just trying to, like, oh, I'm I'm fairly technical. Now I can build stuff.

Mhmm. So I feel like they've gotten pretty good in that regard.

Yeah. Like I said, it's definitely my bias talking and the fact that I, like, almost don't want them to be good. So Yeah. Yeah. I think I'm aware of that at least. That's that's actually a funny a funny question that I've I've asked myself a lot of times JS that, like, would you go back

to a point where we didn't have any of this AI stuff for coding? And I I would probably say, yeah. I would Sure. Like, I was Well, let's go to your happy. Yeah. I was I was skilled. Very happy with where we were. I enjoy a lot of these new tools as Wes. And I Node love using them, and I think it's great. And and I'm not, like, pooh pooh ing them at all, but I also would be totally happy if stuff ever popped up either.

Yeah. I will say that I've, I think if you use the tools correctly, not that we need to go in on this, but, like Yeah. I feel like I've learned a lot more this year than I have in years past because of these tools. No kidding.

They can unstick you a lot. Yeah. Yeah. That's true. And they can put you in spots where you might not have been able to do things. You know? Like, I did several hardware hardware projects this year where I probably would not have been able to to do that. Right? You know, writing a little bit of c plus plus or just simply knowing where Wes to go and what tools to use and to go next. I don't know that it would have been able to do that without an AI assistant sort of pointing me in the right direction. The next one we have here is framework choice matters less.

So AI tools will get better at ingesting documentation, understanding which version of a library you're using, React versus Svelte or whatever.

I don't know. I put a question mark behind this one because I'm not sure if framework matters less. What do you think? I don't know if framework matters less. I do feel like the

other parts of this got absolutely correct. The tools got so much better. Opus 4.5 can write perfect Svelte five code every single time.

No back and forth whatsoever.

And that's because of, one, the model got better, but two, also, there's, even before the model got better, there was things like the MCP servers that can do auto like, there's so much better ways of interacting with the AI. I think that it did, but whether or not this one was lagging behind, funny enough, I did not look at our predictions review before writing our predictions for next year, and I wrote word for word, framework choice will matter less because of AI. And now I'm looking at this. I'm like, oh, man.

So I probably wrote this one in the first place because this is the something that I I I hope I hope that we can pick the tools that are best for the end result rather than the tools that are the most popular.

This is me. You know? So I don't I don't know. Maybe this one,

was a a fail in terms of that. It framework choice probably does matter less, but I don't think people have caught up to that. No. No. I don't like, I think it's gotten a bit better with context seven, MCP, but I still don't feel like it's there. Like, I've I've been trying to use this JavaScript framework called Manifold for doing three d renders.

It's similar to, like, something called OpenSCAD in the three d space where you can just, like, create things and merge them and extrude them like a three d program you would for printing and whatever. And it just I want it's so much better, but it just AI sucks at it. And it's fully typed in all of this, and I just cannot get it to to do better. So I should it's probably been about six months since I've played with it. I should bring it back now and see. Bring it back. Cloud Node, Opus 4.5.

I think you might be surprised.

Docs have no examples. It's simply just here are the interfaces.

Here are the methods. Mhmm. Here are the values that are returned, and it just cannot fathom how to use it. It just keeps imagining things. Next one, web components and standard stack, Wes awesome takes off. Man, this one did not happen. And let me say, I

like Web Awesome, but it did not take off. It's not even out of alpha yet, I don't think. It might be in beta now. I would be interested to know what's taking so long.

Maybe it JS. Our latest release is version three point zero, and it looks like it was six. It was released.

Yeah. You know? I I use Web Awesome on I will say on Synhax, I used Web Awesome.

And, for the most part, it was a really nice experience.

So there was another really good one, quiet UI from Corey Levitzka, who was a part of or is a part of Web Awesome, but it is no more.

Oh, man.

Yes. Happened.

Kawhi UI was from Corey who built the Web Awesome as well.

Yes. And I know what happened here. I'm not at liberty to say on the podcast, but it's a shame because Quiet UI did look fantastic.

It does seem like somebody forked it, but I don't see any progress on that either. So, maybe you can still get the forked version of it or maybe not. I I'm this is something that I was really excited to see and then really bummed out to see is currently unavailable due to some complications.

But yeah. So I think developers are using web components more. I do think web components have gotten better. I think they have grown in popularity.

I personally use them more. People won't author their own, but will happily consume them. I think all of those things happened, but Wes awesome did not take off. Prediction.

Yeah. Here's my my take on all of this new and better stuff because we see it in CSS as well.

CSS has had all kinds of new stuff in the last year. A whole bunch of new stuff is coming up, and you don't see a lot of people using it. You still see them using whole bunch of JavaScript stuff. And I think the reason behind that is because people just want to use, like, a Shad cn or they wanna npm install something and and have it just work, and they wanna be able to debug it easily and whatever.

And they don't necessarily care that there's all this, like, standardization and beautiful fundamentals behind it. But I believe that is going to change because I think that a lot of these UI frameworks are going to be using them. Like, obviously, Web Awesome is is already behind that, but I think that they will become more mainstream. And I'm gonna stop predicting for the next year because I'm gonna put this in the next episode, but I think that's why. Next one, AI browsers and search Copilot style workspace style tools become more normal. So the way that we thought about this is, like, what does coding actually look like? Right? You have sitting in an editor, you have this, like, new agent UI that all a lot of these tools are trying to push on you. Like, cursor just by default now goes to the agent tab instead of the code tab, which is kind of obnoxious.

And all the other tools have something similar to this, and, GitHub is going absolutely ham on trying to shove all of the stuff into GitHub and down your throat. I don't know. Like, are these becoming as as mainstream as we thought they would? Right? I I certainly find them to be useful Wes you you can go on GitHub and say, hey. Review this or grab this grab this thing, but I don't know that that's the UI that is the absolute killer UI. What do you think? I think that

they have become more reliable and more mainstream. I do think that has happened. And the the latest stuff where you can assign different agents to different things that they announced the GitHub universe, all that stuff is is great. So I I personally think that, yes, these things did become normal and more mainstream.

Yes. I will say that. Have they become great? I don't know. I've used them to fix some issues here and there. Am I using it to do most of my coding? No. Next one we have is the AI browser will become inevitable, and I'll I'll throw the next one in there as well, which is OpenAI

will launch a browser. So OpenAI obviously shipped their own browser. It Wes absolutely awful. Very slow.

I don't know anybody that's using it. It's very slow, very limiting, very frustrating to use, but they're trying to cram you into the hole that they want the web to be. Plus, like, what, Scott changed browsers, what, six or seven times?

Maybe more than that. Yeah. What what browsers did you use this year? Let's let's count them. Like, for longer than a week? No. Just, like, what did you what did you dip into? There Wes, obviously, the OpenAI Browser, which is I don't even what is it called? Comet? I'm gonna I Node. That's that's perplexities. I'm gonna be real with you. I did not ever download the OpenAI Brow OpenAI Browser, not even for half a second. But I did use Dia AI Browser. I did use Perplexity's Comet for a good long time AI Browser. I used Yarn for a long time. I used Chrome Chrome Canary.

I tried to use Firefox for just a little bit, and then I've been using Helium for the past month and a half or so. And really liking Helium is where I'm sticking for now. I will say with per the Comet browser is great. There's a lot of I didn't wanna use it for my daily browsing, because of privacy concerns. They're sending a lot of information to perplexity.

And I just I don't want them to have access to my bank account and all this information. But, like, when I wanted to do a little bit of things here and there, it was super useful. Like, for instance, if I wanted it to put me on like, CJ was flying to New York. I wanted to be on CJ's flight. I copied his flight number. I pasted ESLint and I said, put me on this flight. It got me all the way to the checkout screen. Now I did just go ahead and put myself on that flight, and I clicked buy and then later realized that CJ is CJ likes early flights. That flight was at six in the morning, and I'll never forgive myself for just yellowing onto that flight because that was that was crazy. I I've never flown that early or at least I have, but not I don't like to. I'll tell you that.

But, yeah, I got gonna be the new Yeah. Oops. Autofill did it. Like, anytime I I sell stickers or whatever or the Syntax Wes store, you get all these, like, emails being like, oops.

I got my address wrong because autofill did it. Now it's gonna be, oops. AI accidentally put me on a flight. Oops. AI put me on a flight. Yeah. And if you wanna know, like, the type of brain rot that you're getting from these AI systems, I didn't know that the flight was at 6AM until the week before I was getting on the flight. And I was like, oh, let me show you what time my flight. Oh, no. What happened here? You know what's one

one tool that I I think which had GPT is starting to get this, but I just want, like, an AI brain where I can just drop things in that I need to remember at some point. Mhmm. You know? Sometimes I'm putting something away, and I say to myself, I have to remember that I put this here because I'm gonna be, like, in in six months, I'm gonna or, like, my wife couldn't find the stockings.

And she's like, I put them away somewhere, and I can't remember where. And wouldn't it be so nice just to have a little, like, walkie talkie? I'm putting the stockings near the kids' boots. You know? Just I want that. No. Just little reminders, you know, or on because I on my watch. You know? Remind me x, y, and z. I use that a Scott. But couldn't I have, like, a brain where it's just, like, remind me and or here's what I'm doing or, I use this type of Scott, like, the lug nuts on the tractor at the cottage. That's something I have written down somewhere, and I needed it at one ESLint. I couldn't find it. And I'm like, I know I put this somewhere or, blinds. I have every window in our house measured up.

I don't know where it is. Is it in a note? Is it in a Google Sheet? Did I put it in a markdown file somewhere on my computer? I don't know. And search sucks on my computer, so I can't find it.

The one I would like is my my, electrical box. I have repeatedly toggled them all on and all off. And every single time Wes I'm doing it, I'm like, oh, when I figure this one out, I'm going to use the marker and write down what this circuit is. And then I find it, and then I do my project, and then I turn it all back on. I say, oh, but oh, okay. Let's go do the next thing, and I never write it down. And, Yeah. The next time, I do that as well. Half of ours are wrong for some reason. Yeah. Yeah. And everybody does it in permanent marker on there, so it's not like you can, who knows what the last owner wrote on there or something.

Alright. Let's talk about CSS predictions.

Relative color syntax. This is where you can take a color, whether it's an HSL, RGB, hex, whatever, and you can use from, and then you can split it out into RGB values, HSL values, and you can increment, decrement them, whatever.

Beautiful syntax for being able to calculate a color.

And, Scott, you said you wanted that 94% usage.

It has been in every browser. It actually reached baseline twenty twenty four, meaning that it's been in every browser for a full year now, and I think that it's very safe to use.

I love relative color syntax. This might be my MVP for the year. Stamp it. I love this API.

I I've built color systems that respond to light and dark mode, transparency.

I built the crazy color contrast automatic color contrast thing the other day with this syntax.

Man, this syntax is underutilized and just all around fantastic. It JS greatly improved my CSS, in just ways that I cannot express on this podcast. Yeah. I love relative color. Next one is vanilla CSS will make a comeback with all the CSS superpowers scoping, anchors, relative color. More teams will just ship plain CSS. I do think this is happening whether or not it has become, like, a movement or what. I I do think this is happening, though, all the time. I do see more and more people shipping some of this latest stuff. I I personally don't even add post CSS to anything. I don't add CSS to anything. I wrote my own utility framework. I wrote, you know, utility framework. You can tell I wrote that because that's me. That's the type of thing I would say.

Yeah. I I I think, man, CSS is just so capable now and is only getting more capable. So I I do think it is making a comeback to just use this. Whether or not it is CSS huge. I think it like, we're still early in a lot of these features, whether it's people still need to support

older, like, iPads. You know? Like, that's probably the biggest. Talk about IE back in the day.

The biggest pain in in support's butt is old iPads. People don't upgrade their iPads or they simply cannot. You know? Like, everybody has an iPad that's nine years old, and they use it every now and then. And they're just they're simply at a browser where they cannot upgrade any further, which is really frustrating.

Yeah. But I feel like we're right on the precipice

Yeah. To this. If they were to install the latest Mac or iPadOS on that thing, it would literally melt their device because it's so bad.

It's like oh. And if you want to see all of the errors in your application, you'll want to check out Sentry at sentry.io/syntax.

You don't want a production application out there that, well, you have no visibility into in case something is blowing up, and you might not even know it. So head on over to century.io/syntax.

Again, we've been using this tool for a long time, and it totally rules. Alright.

CSS mix ins and function Vercel solidify. So there Wes rumblings at this time last year of adding both mix ins and functions to CSS.

So similar to if you've used Sass before Wes you can, apply a whole bunch of CSS to a selector, or you can make custom functions that will be able to take inputs and and return values. So the CSS mix ins is now a solidified spec. It's still in draft mode, but it's been updated as much as, what, about a month ago as of recording.

And the spec has has gone forward with only defining custom functions as part of one. So, then the idea of mix ins, which I I'm pretty sure I believe we did all show on it, but I'm pretty sure mix ins is being able to apply multiple functions to one.

Is that right? I, I'll be real with you. Even as far as going back as SAS goes, I never had a firm grasp on difference in usage between mix ins and functions because at the end of the day, you're just looping in a bunch of complex logic. I guess functions are for defining values,

and mix ins are defining chunks of CSS, or am I is that Yeah. I'm just reading through a spec really quickly, and that's that's what it JS. Is functions are for calculating a single value, and a mix in is for returning multiple Okay.

Values from that. So if you if you wanted to have, like, a mix in to make, like, a a syntax button, you know, or it has has a specific border radius and a drop shadow and a background and calculate the that, that's where a mix in would work. So in order to have those things, first, you need functions to be able to actually, take inputs and then return values, and then you can get more complex with it. So it looks like, the spec has been firmed up. And it looks like in next year, we're going to start seeing this

land in browsers. I'm pretty excited about that. Yes. Yes. I am very stoked for this. I'm gonna use this all the time. There's so many things I I would encapsulate in a little function instead of having to write some massive amount of calc and who knows what inside of my CSS. So true. Yeah. Just be able to put that because those things are so hard to reason about and just being able to put that in a nice function, which is simple,

is going to be key. Yeah. But on the same regard, we also got we also predicted that container queries are going to be shipping everywhere. Sorry. Not container queries. Container style queries. Container queries have been in the browser for for a couple years now. Style queries, which is you'd be able to essentially and Wes essentially, it was if statements against container properties. So you can say, like, if this container has a flag of dark or if this, container has a variable that's set to subtle, then do the following things. So we got that. That shipped in all of the browsers. However, we also now have if statements added, to the browser. It's only in Chrome right now. It's limited availability, but the idea of if statements are not just applied just to container queries, but you can also apply them to to media queries and what else? Feature queries. So that's still a ways out,

but I'm pretty excited about it. Yeah. Yeah. I I'm I'm stoked about it too. There is a a point, though, when you when you wonder, like, what can be done just by adding classes or not or whatever. And the if statement part is very, very interesting. I I did a little bit of a deep dive on it at some point. So, yeah, it is a cool cool API, container style queries. You said the style query is shipped in every major browser?

Yep. Let's look at the can I use for that?

No.

I'm seeing that they are partial support. Yeah. Partial support refers to only working on CSS custom properties in a style query. So it only works But not in Firefox at all. Oh, you're right.

Why did I see that? Okay. Yep. Wrong about that. Yeah. Well, not in Firefox, unfortunately.

But, honestly, I think that all of the stuff needs to land before we, like, fully see it. You know? So we need if statements Yeah. Style queries, functions, mix ins, all of the essentially, CSS is becoming a programming language,

and then we're gonna be able to do some serious, serious stuff with this. Yeah. Yeah. I I was gonna say that would be low key a groundbreaking or ground shifting, API to have dropped in all major browsers. I'm stoked for, style queries. Vertical centering jokes will stubbornly persist. Nailed it. Yes. People still make jokes about this stuff. Man, it's so easy to center anything vertically, horizontally, whatever. There's billion different one liners to do it. It's not hard, folks.

So You can vertically align something on a block level element, which means a div. You don't use grid or flexbox or anything with the align content center.

You don't have to say that joke anymore.

Wes. Stop doing that joke. Although it doesn't work in CSS battles, because every single time I try to use it in a CSS battle, it fails because of however their renderer is doing. There's just something weird. Yeah.

Yeah. Editors, IDEs, and dev tools. So are we predicted Versus Node will reach feature parity with cursor? I don't know that that happened. So the agent agentic stuff in in Versus Code got very good.

However, I feel like the and I know that the tab completions, which is one of my favorite AI coding things, is just not there. Like and I've been I've been DMing with a bunch of people that work on on Versus Node and being like, I've sent them videos, and they turn those videos into like, I have I have code and cursor up, and I was like, these are the four things I wanna do. Yeah. And I I just the cursor does it beautifully, in tabs. It does exactly what I want. It jumps to where I want, and then Versus Code, like, forgets, forgets a closing curly bracket, which is so frustrating, and it's it just it's not quite there.

They're obviously working on it, but just cursor is still miles ahead on the tab completion. So I I don't think we got that one.

It's gotten a lot better, though. I'll tell you that. Versus Node gotten a lot better. Yeah. I I think the tab completion is not something I use that much.

I'm gonna be honest with you. I don't. Yeah. I don't I don't know if it's because Zed doesn't have as good of tab completion as cursor, and I used Versus Node primarily before moving to Zed. And so, like, maybe I just haven't had the really good tab completion

in a little while. Yeah. That's the thing. It's like, I've got so used to having very good tab completion in cursor, but then you you talk to people about that work on these things, and they say it's really hard because some people think it's too eager.

Some people think it's not eager enough. Some people hate it. Some people can't live without it. And it's all of this, like, everybody's just kinda at a different spot with it. Like, even sometimes I have to turn it off when I'm doing some types of code. I have to turn it off because it's just way too aggressive as to what I want. It's like, I I maybe can predict everything you wanna do here, and then it's like, no. No. I'm just trying to put a border on this thing at all. Please. Wes. Chill out. But then other times, I'm like, well, yeah.

Keep going. I want you to do it. So I don't know. Like, I need I almost need, like, a knob where I can, like

Yeah. Yeah.

Yeah. How how eager should this be? Yes. Much but, like, yeah, that's it's a tricky one. More Versus Node forks will appear, of course. And we had anti Sanity, and Google had two forks of their own in 2025 pop up. They have one called JUULS, and they have anti gravity, which is kind of like Wind Surf Yes. Which is Wind Surf, but also not in, like yeah. It's kinda their launch kinda flopped, though, because they they launched it, and then nobody could use it. It didn't work for, like, a like, a week.

And it's still kind of. Yeah. There's still reviews.

Yeah. So it's unfortunately, I think they pushed it out way too early because everybody tried it, sucked, and everyone went back to Cursor,

which is really frustrating for them. And they got duels too. Like, what are they doing? I it's classic Google stuff, but you'd think by now that they would they would figure this out. Yeah. And they have IDX, which is now Firebase Studio. So Yeah. Google has at least three

IDEs agentic IDEs.

The forks will continue until more morale improves.

Next one here is for React ecosystem.

React compiler

drops Babel. We said that because Babel is ancient technology at this point, that the React compiler would drop it. No. That's not true. It did not happen. Scott happen. Is anybody using the React compiler? I'm not That's a good question. Yeah. As as much as people talk about it and as much as people say, like, rendering performance and and whatever, and there are lots of situations you can point to where, like, this thing is slow.

Like, GitHub pull request is a really good example.

I think most, like, relatively simple React apps don't have that problem, or people don't even know that they are rerendering unnecessarily.

I think it's people don't Node, not that they don't have that problem. Yeah. They don't know what they know.

But then we also had one here JS React Vercel components will pop. They'll hit their stride, and we'll see a lot of universal RSC components. Meaning that, like, my my prediction here was that you'll see a people ship React server components that include the client logic, the server logic, the styles, API calls, all of the stuff bundled into one component that you can Node. And I will argue that that did not happen.

Yeah. And in fact, we just had a mass of security issues with React to Shell, which was basically every React website that used server components. So not every React website, but every every React website that used server components. My own website was on Wacu.

Obviously, all the Next. Js websites out there, anything that's built on Vercel with React server components, they were all effective. And it was a pretty big pretty big,

mobility. Yeah. Remote execution, remote code execution with just a post request. So, I mean, it was pretty yeah. So we're Pretty rough, which is honestly, I don't

I think that will sour a lot of people on React for quite a while. And it sucks, And it's I don't wanna, like, point fingers, and I don't think anyone should be dancing on the grave, because that this happened. I know a lot of people like, it was very serious, and I'm sure a lot of people had a very bad week.

That sucks. And I think that that will be a a bad band aid. People are waiting for something like this to happen to React, you know, because of all the crazy APIs in React Vercel components.

Yeah. So Yeah. Bit of a bummer.

Yeah. A bit of a bummer there. Next one, Remix emerges as something new.

That happened. Remix Yep. Is emerged as something new. Whether or not that's being used by anybody, I don't know. That's not out yet.

It's not out yet. I was not at Remix Conf. I don't know the timeline here. Oh, no. You should go back and listen to our episode on,

the Remix.

I forgot. Wes stuff enters my brain and just exit. If I'm like, I'm not I Wes, I'm not picking up remix. I I like, I don't got room for it.

Announce what it is going to be, and it has nothing to do with React anymore. Yep. And it is a sort of a full stack entirely JavaScript. It still uses JSX.

It's all based on web standards.

We guess it might be a Shopify related product. That is not the case.

We guess that it might be a rack server components based headless commerce thing. That was not the case at all.

It has nothing to do with React anymore.

And yeah.

Next one is React Native will have its time. I I React Native is is kind of its own thing. Like, it gets used a ton where it's Node, but where people don't need to build native apps, they're obviously not going to be using it. So I don't know if we nail that one either.

Yeah. I I think it has. I think React Native continues to get better. Expo continues to get better.

The more I see from React and one thing React Native, I think, just added CSS, better CSS support for a CSS grid or something. They have been been been crushing. Yeah. I think they also felt the hype around links. I follow a number of React Native folks on Twitter, and I do see kind of constant hype around various things in the React Native ecosystem. Links

for anyone listening, people are probably gonna try to Google this. It's l y n x. This is a React Native competitor from ByteDance, which is the owner of TikTok.

And they and they launched their own version of React Native, which powers a lot of the big, big apps that they have in China.

Yeah. Lynx is really cool. I'll tell you that. Keep your eye on Lynx.

Next one, TanStack stark, and TanStack will pop. Yes. We nailed this one. Tanstack has got hype around it nonstop. All the Tanstack stuff, I've been seeing major things Node to Tanstack.

And I I think that they have a lot of it it's funny. I mean, the Tanstack ecosystem Vercel has, in my mind, legitimized quite a bit, but also Yeah. Grown quite a bit. And then They just launched TanStack AI, which I think is gonna be huge.

Which we knew about early, by the way, folks. Yes. Because yes. That's what Tanner told us. After the episode we did with him. He told us what it was gonna be.

That's pretty cool. So just like a standardized AI SDK that can be used in JavaScript, Pnpm, and what else is it? I think Python, which is really interesting. He they're going with, like, every language you can use this SDK in, not just like TanStack is known for like, yeah, it works with React, but it also works with Svelte, and it also works with Vanilla. But now he's going outside of that and just working with every single language, which I think is the move because then you can if we standardize on these ideas of, like, responses and Wes

and the whole thing is gonna be beautifully typed, I'm pretty excited about that. Yeah. I'm gonna do some tutorial content on this Tansa AI stuff. Really stoked about it.

I had some SvelteKit stuff. So SvelteKit gets more granular data loading. They did. They did not get islands or they did get some kind of hydration control. They didn't get they didn't get the island stuff, but they did get more granular data loading with remote functions, and they're they're dope. So that is component level data loading. Yes. I have some local first apps will take off, and I do think they have had quite a bit more hype behind them. I don't think it's reached critical mass where people really understand the capabilities of local data caching and how how great that is. And then, you know, when that happens, everyone's just gonna act like it's some brand new thing. But I I I think that is still to happen. But, man, local data has definitely taken off. I think more people know what the word local first is even if they're using the term incorrectly.

But I do think most people do understand that local data, is a viable strategy.

Yeah.

For BUN, we had two predictions. We'll say BUN will keep doing wild but loved nonstandard features, and they certainly continued that. They they doubled down on their Node this year, but they also did a lot of nonstandard features, which was they essentially became a Vite competitor, and they also implemented their own router. So Bun is is an awesome tool. But probably the big one Wes said here is Bun will launch a platform as a service. So we're like, well, Bun is VC funded. They have to make money at some point.

What are they gonna do? You know? The the clear path forward is just, like, they're gonna release a hosting environment.

That did not happen. They got acquired by Claude because Bun is powering, like, Claude Node. Yeah. Scott it. Sorry. They got him acquired by Anthropic.

And that's a pretty interesting move because JavaScript as a almost like a compiled executable I know it's not actually that, and it's much bigger. Last time I said compiled JavaScript, people, like, lost their minds on me.

But being able to take a fast and easy to code core, which is BUN, and put it into products like Cloud Code is is why they've they've acquired them and and done it. So I don't know why they did just use Cloud Code to spin up a a bun, themselves. You know? It's gonna take all the coding jobs, but I guess they had to pay a whole bunch of money for some of the smartest people in JavaScript to to do that. I don't think, I don't think AI is able to write Zig because Yeah.

Zig just moved off of GitHub, so now, their Node isn't getting flipped up automatically. Right? Next one, Vite will stay king. Vite remains the dominant bundler server choice. Yes. It is. I still firmly believe that, man, Vite rules Vite powers so much, and anything hits Scott Next. Js is Vite. That's really where it's at. Or, I guess, BUN, but yeah. Rolldown.

Rolldown JS, man, rolldown was released, but it's not the default yet in Vite. In rolldown is It was the default in seven.

What? Was it not?

I don't know. I'm so there's too many things, Wes.

Yeah. I'm pretty sure I was upgrading something the other day, and I went through the breaking changes.

Roll down.

How to try roll down? You have to use Vite npm roll down Vite latest to use roll down.

The current docs say that Vite is planning to integrate roll down.

To use roll down, you have to use roll down Vite. It is still considered experimental.

Yeah. You're right. Seven point o, they rolled it out, and you have to opt in. It's eight point o, which is in beta one as of today.

No.

Where it will default to roll down Vite. So probably in a couple months, we'll see that be the default. I've been using it on all of my projects, and it's been absolutely awesome. Yeah. It's great. Yeah. There's also Node thing that we missed on this is Vite plus.

So Vite plus was what is it? Give give me the rundown. Oh, man. I'm I'm gonna be honest. I don't have a good answer to that question.

The whole idea behind Vite plus is not just not being like, the dev tool and and bundler and builder that it is right now, but it also is your your linter, your formatter, your test runner, UI testing, all of basically, absolutely everything, and and that's their their way to make money. Right? They they raised a whole bunch of money. Whether that will be the case or not, we will see, but I don't know. Part of me is just like, yes. Please just give me something that does everything in one.

And then the other part of me is just like, no. I wanna be able to pick my tools and and maintain this sharded JavaScript landscape that I I've grown to love.

So we'll see.

I wanna use Vite for everything, and I already have only been using Vite for everything. And As as long as it stays amazing, you know, and doesn't and and they've proven that to me over, what, eight versions Wes I've upgraded happily every single time without issue. And as long as it just keeps getting better and I'm not bogged down by anything, then, yes, I think that's the move. And the v team has always delivered

consistently. So

yeah. We said Laravel's gonna release a CMS. That did not happen. They said they're gonna acquire a or tightly partner with Statimic because at the time, this was during all of the WordPress drama. Mhmm. And everybody was like, well, I'm gonna go on a new CMS. All of that WordPress drama and whatever, like, not to downplay it, but, like, I don't think that a lot of people moved off of it. The people that use WordPress love it, and I there's still not a very good replacement for what WordPress is, which is a huge ecosystem with lots of plug ins and lots of stuff in it. And it is just this whole monster of its own that I don't know will ever be entirely replaced. There's been bytes taken out of it. You know? You've Scott, like I think the biggest bites are, like, things like Squarespace and Webflow and whatever. They've taken bites out of that, but it's still so huge.

I don't think that the majority of WordPress users even know what Yeah. Anything in regards to any of the drama even happened or was. So, developers, sure. But, like, for users, no.

Yeah. Yeah. I agree. That's our predictions. Tallying it up.

I think we got a a healthy 87%

accuracy. I think we did pretty good on that. What do you think, Scott? I think we got a 100% accuracy because that's what I choose to believe. That's my reality.

Yes.

Let us know down in the comments what you think we hit and what we didn't, and, of course, we will see you one week from today on our episode talking about our twenty twenty six predictions.

Yeah. Do you wanna do a sick pick before we get out of here?

Yes.

I have a sick pick that I just remembered the other day, and this is not a very good winter pick, but I don't care. It's this is something I went to your house, and I slept in your guest room, and you had a woozy fan. Yeah. The woozy. This fan that goes it a fan that goes on two axes.

Mhmm. And it's tiny, and it's nice and quiet, and it has my favorite feature error, which is you can turn the freaking blue LEDs off while you're sleeping.

So we got one of the little ones, and then we got another one that's on, like, a pedestal. And they're so compact, but they still they put out, man.

They put out, and it's, like, the best fan I've ever had, and I've been so happy with it. So next year, once you're getting a little toasty and you're like, we need a new fan, go to Costco or whatever and grab the Wuzu fan. They're unbelievable.

And I'm a I'm a fan noise kinda guy, so I like a little fan on at all times in the when I'm especially when I'm born to school. Yeah. Yeah. I I have, like, a little USB

sound machine that I travel with because I just I don't like when it's dead quiet. I like a little bit of

Yes. On there. Yeah. Same. I'm gonna sick pick a little UV black light. It's like a little almost like a little flashlight black light. I got it I got it from Why do you hold it like you're a cop? Yeah. That's actually a good question.

That is how I hold it when I use it, though.

Yeah. Yeah. No. I I've been using it actually, well, I've been using it to clean clean, like, in the kitchen or something like that. I'm I'm like, oh, man. You see everything with this. I actually got it for my skin, believe it or not. But, like So wait. Wait. What does it do? It's a basically a really high powered black light.

And you see what? You you see every little crumb ever. You can find all kinds of stuff. I lost a pill last night, and I was like, I Node it was here. And then I was like, oh, let me bust out the black light. I turned it on, and that pill just blew, like, bright white. It was like, I found it instantly, and it's so good for finding little stuff like that. Or, like, our daughter had drawn on the carpet with, like, some kind of highlighter or something, and we could barely see it. I turn on that that black light, and all of a sudden, it's, like, right there. Now we clean it better. Like, it is I've been finding so many little use cases for this thing. And it's $9, this little, like, black light flashlight. I've been using it to all do all kinds of just, like, cleaning around the house or whatever.

Or, like, when I'm, you know, cleaning the kitchen at night, I'm like, I'm not like a germaphobe or a clean freak or something, but I turned on. I was just like, oh, there's all these crumbs here that I missed or whatever that you're just not not seeing otherwise. So, it it it's funny. I Scott it for a completely unrelated use case, and now I'm just using it for all kinds of stuff. So Wow.

Terrifying.

Terrifying. It's not terrifying. Also, when you pointed at the the glow in the dark syntax sticker Oh, yeah. It instantly turns neon green. It's pretty sick. So

awesome.

Yeah. So that's it for our predictions. Let us know how you did on your predictions for this year, whatever you thought was gonna happen. Let us know Wes something that you thought was gonna happen this year and didn't happen, because that's what we're interested in. And leave your predictions for the next year as well because we're about to hit you up next week with what we think is gonna happen in 2026.