966: 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 predictions. So every single year, what we do is we look at where web development is and we say these are our predictions for the next year. What's going to happen? What's going to flop? What's going to pop? And we are going to 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 going to happen. So my name is Wes Boss. Happy holidays. Merry Christmas, everybody. Hopefully you're enjoying it. This is going to be a really fun episode. I'm here with Scott Talinsky. 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'm just genius. Yeah, I just have an incredible intuition and all of that. Oh, yeah. No, that's great. I like that a lot. Did it flop or did it not? That's a great metric I think we can all live by here. Did it flop or did it pop? Flop or pop? I love these episodes because this is the thing I like doing the most. Just pontificating on, like, theoretical who knows what. Is this going to happen or not? I feel like I'm always a little too optimistic on stuff. So, very interesting. I have to say, last year we had a lot of, like, man, we were a little too optimistic on that. But I have to say that this year 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. I don't know if anybody's ever said what's wrong with the date API. I don't know. Fair, fair. So the temporal API will allow you to encapsulate durations. So if you've ever wanted to say, I want to somehow make three days, how can you say that in a standardized way? Or if I want to do January 25th to January 26th. 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 is going to 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 26th, it was added to Firefox. Good. And December 3rd, it entered into Chrome beta, which means that a month later, January 7, we're going to January 7, 2026, we're going to be it's going to 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 years. 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 a polyfill for like three years now. So, you know, I've been using it for a while and I've shipped stuff with that polyfill. So, yeah, I'm stoked about it. But I this is the type of movement that it feels like I've been wanting to see for a long time. And yeah, we're at. Yeah. You can start to learn temporal. y'all. 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 going to say we missed 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 think as people start to say, what does it mean when you don't have to pay anything for it, when it just runs on your user's device? You could run it constantly in a loop. So I don't think we're quite there just yet. You are seeing a little bit more niche models. Toxicity detection is a really good one where if you need to detect if something is spam or 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 um they're really just playing with a lot of that right now same way transformers js so i'm gonna punt that next year you say miss well like but listen i'm saying that as 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 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 so i think we saved ourself with the wording here i was talking to some of the um 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 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 model. Plus, they're way better. And unfortunately, I don't think a lot of people necessarily care that much about the privacy aspect, which kind of pains me to say. But I think that's just the reality. Yeah. Well, the next one is that WebGPU unlocks fast local machine learning. And I do not – I think it might be a little bit early, man, but TypeGPU is doing some cool stuff where they even have like an example of text prediction and stuff using WebGPU. 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 going to make it fast in browser. Like I have an example where I generated alt text with Transformers.js, and the WASM example versus the WebGPU example is six times faster. And then it moves from a cool demo to like, oh, we could actually use this. So I think that's the one that we're going to hopefully see more of in the next year. But that's the next show. I'm not going to 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 models will plateau in quality, and we're going to see big improvements on how the products use the existing models. And I think we nailed that. But let's dive into the different aspects of where are we with AI versus 365 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. 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 Anthropic? Yeah. It is Sonnet 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 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, like the quality of code that agents are able 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 co-pilot 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 was 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 i i'm attributing that just to tools but maybe i'm wrong about that 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 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 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'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 models as 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 um there obviously are some use cases for it like we're designing 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 sorrow was funny for like a week and a half and haha and now it's just like 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 oh i made this now haha it's like oh it was really funny when i made west boss in the sopranos looking like pikachu okay okay that's that's very funny to me but they're but sora is that was it so funny though like the one where the the a pig comes through the window and the man's sitting on a toilet oh man i still laugh about that like i don't know there's something but it's like it's 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 all the energy and stuff that used 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 going to come out sloppy uh because they don't have an artistic vision for it yeah you don't have crafts people using this stuff it's the marketers or the grifters doing it that was actually one thing about sora is that sora somehow figured out how to make the videos funny but then Like after a while, you kind of understood all the hilariousness, you know, and it only knows how to – 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 of weeks, you realize, oh, that's just the recipe for making it funny. Yeah. Next one. Text to UI tools get really good. Things like Bolt.new, v0, et cetera 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 say really good they got better but that my bias talking i think we we nailed this prediction because cursor vs code anti chrome mcp puppeteer all of these agents now have the ability to take the ui into account both like screenshotting it both taking videos as 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 v0 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 non-dev space the tinkerer space you know and that's that's a whole new aspect of people who are just trying to like, oh, I'm fairly technical. Now I can build stuff. 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 almost don't want them to be good. Yeah. I think I'm aware of that at least. That's actually a funny question that I've asked myself a lot of times is that like, would you go back to a point where we didn't have any of this AI stuff for coding? and i would probably say yeah i was like i was well that's very happy yeah i was i was skilled very happy with where we were i enjoy a lot of these new tools as well and i know i love using them and i think it's great and i'm not like pooing them at all but i also would be totally happy if the stuff ever popped up either yeah i will say that i've uh 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 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 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++, or just simply knowing where to go and what tools to use and to go next. I don't know that I 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 5 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. But 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 something that 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 know. Maybe this one was a fail in terms of that framework choice probably does matter less, but I don't think people have caught up to that. No, no. Like I think it's gotten a bit better with Contact 7, MCP, but I still don't feel like it's there. Like I've been trying to use this JavaScript framework called Manifold for doing 3D renders. It's similar to like something called OpenSCAD in the 3D space where you can just like create things and merge them and extrude them like a 3D program you would for printing and whatever. And it just – 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 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 code, Opus 4.5. I think you might be surprised. Because the docs have no examples. It's simply just here are the interfaces. Here are the methods. 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 in standard stack. Web 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 is. Our latest release is version 3.0. And it looks like it is stable. Yeah, November 6th it was released. Yeah, you know, I use Web Awesome on, I will say, on Sinhacks, I used Web Awesome. And for the most part, it was a really nice experience. So there was another really good one, QuietUI, from Corey LaVisca, who was a part of or is a part of Web Awesome. But it is no more. Oh, man. Yes. What happened? QuietUI was from Corey, who built 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. 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 web awesome did not take off yeah 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 a 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 want to npm install something and have it just work, and they want to 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 going to stop predicting for the next year because i'm going to put this in the next episode but yeah 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 uh 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 mainstream as we thought they would, right? I certainly find them to be useful where 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 latest stuff where you can assign different agents to different things that they announced at GitHub Universe, all that stuff is great. So 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 throw the next one in there as well, which is OpenAI will launch a browser. So OpenAI obviously shipped their own browser. It was 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 browsers did you use this year? Let's count them. Like for longer than a week? No, just like what did you dip into? There was obviously the OpenAI browser, which is, I don't even, what is it called, Comet? I'm going to, no, that's perplexities. I'm going to be real with you. I did not ever download the 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 Arc 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 the Comet browser is great. There's a lot of I didn't want to 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 into perplexity 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 um but yeah that's gonna be the new oops autofill did it like anytime i i sell stickers or whatever or the syntax swag store you get all these like emails being like oops i got my address wrong because autofill did it now it's going to be oops ai accidentally put me on a flight oops ai put me on a flight yeah and if you want to know like the type of brain rot that you're getting from these ai systems i didn't know that the flight was at 6 a.m until the week before i was getting on the flight and i was like oh let me see what time i fly oh no what happened here you know what's one one tool that i i think chat gpt is starting to get this but i just want like a an ai brain where i can just drop things in that i need to remember at some point 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 lot but couldn't i have like a brain where it's just like remind me and or here's what i'm doing or um i use this type of screw like the lug nuts on the tractor at the cottage that's something i have written down somewhere and i needed it at one point i couldn't find it and i'm like i know i put this somewhere or um 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 that 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 it's my my uh electrical box i have repeatedly toggled them all on and all off and every single time when 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 oh yeah the next time i do that as well half hours are wrong for some reason yeah yeah and everybody does it in permanent marker on there so it's not like you can uh who knows what the last owner wrote on there something all right let's talk about css predictions um 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 2024, 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've built color systems that respond to light and dark mode, transparency. I built a crazy color contrast, automatic color contrast thing the other day with this syntax. Man, this syntax is underutilized and just all around fantastic. It has greatly improved my CSS in just ways that I cannot express on this podcast. Yeah, I love relative color. Our 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 do think this is happening all the time. I do see more and more people shipping some of this latest stuff. 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 no utility framework. You can tell I wrote that because that's me. That's the type of thing I would say. Yeah, I think, man, CSS is just so capable now and is only getting more capable. So I do think it is making a comeback to just use this, whether or not it is ubiquitous. I think we're still early in a lot of these features, whether it's people still need to support older iPads. That's probably the biggest. talk about ie back in the day the biggest pain in supports 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 if they were to install the latest mac or ipad os on that thing it would literally melt their device because it's so bad. And if you want to see all of the errors in your application, you'll want to check out Sentry at sentry.io forward slash 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 sentry.io forward slash syntax. Again, we've been using this tool for a long time and it totally rules. All right. css mixins and function specs solidify so there was rumblings at this time last year of adding both mixins and functions to css so similar to if you've used sass before where 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 return values. So the CSS mixins is now a solidified spec. It's still in draft mode, but it's been updated as much as about a month ago as of recording. And the spec has gone forward with only defining custom functions as part of one. So then the idea of mixins, which I'm pretty sure I believe we did a whole show on it but I'm pretty sure mixins 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 sass goes I never had a firm grasp on difference in usage between mixins 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 mixins are defining chunks of css or am i is that yeah i'm just reading through the spec really quickly and that's that's what it is is functions are for calculating a single value any mixin is for returning multiple okay values from that so if you if you wanted to have like a mixin to make like 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 mixin would work so in order to have those things first you need functions to be able to actually uh take inputs and then return values and then you can get more complex with it so it looks like um 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 going to use this all the time there's so many things 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 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 a couple years now style queries which is you'd be able to essentially and we 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 um to the browser it's only in chrome right now um 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 media queries and what else? Feature queries. So that's still a ways out, but I'm pretty excited about it. Yeah, yeah, I'm stoked about it, too. There is a point, though, 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 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 queries 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 yeah 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 mixins all of the essentially css is becoming a programming language and then we're going to be able to do some serious serious stuff with this yeah yeah i was going to say that would be low-key a groundbreaking or ground shifting uh api to have dropped in all major browsers i'm stoked for uh style queries vertical centering jokes will stubbornly persist nailed it yeah people still make jokes about this stuff man it's so easy to center anything vertically horizontally whatever there's a 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 yes 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 something weird yeah yeah editors ide's and dev tools so are we predicted vs code will reach feature parity with cursor i don't know that that happened so the agent agentic stuff in in vs code got very good um 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 vs code and being like i've sent them videos they turn those videos into like I have code and cursor up, and I was like, these are the four things I want to do. And I just – cursor does it beautifully, and tabs does exactly what I want. It jumps to where I want. And then VS Code forgets a closing curly bracket, which is so frustrating, and it's not quite there. They're obviously working on it, but just cursor is still miles ahead on the tab completion. So I don't think we got that one. it's gotten a lot better though i'll tell you that vs code's gotten a lot better yeah i think the tab completion is not something i use that much i'll 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 vs code 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 and 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 kind of 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 maybe can predict everything you want to do here and then it's like no no i'm just trying to put a border on this one thing at a time please yes 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 how much but like yeah that's it's a tricky one more vs code forks will appear of course we had anti-gravity and google had two forks of their own in 2025 pop-up they have one called jewels and they have anti-gravity which is kind of like windsurf yes which is windsurf but also not and like yeah it's kind of their launch kind of 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 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 they got jewels too like what are they doing 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 authentic 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 uh that the react compiler would drop it no that's not true it did not happen that not happen um 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 um 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 re-rendering unnecessarily i think it's people don't know not that they don't have that problem yeah they don't know what they know um but then we also had one here is react server components will pop they'll hit their stride um 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 do and i will argue that that did not happen yeah and in fact we just had a massive security issue with react to shell um which was basically every react website that used server components so not every rack website but every every react website that used server components my own website was on waku um obviously all the next js websites out there anything that's built on parcel with react server components they were all effective and it was a pretty big pretty big uh it allowed ability yeah remote execution remote code execution with just a post request so i mean it was pretty yeah so 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 want to like point fingers and i don't think anyone should be dancing on the grave um 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 will be a bad bandit people are waiting for something like this to happen to react you know because of all the crazy apis and react server components Yeah. Yeah. Bit of a bummer. Yeah Bit of a bummer there Next one Remix emerges as something new That happened Remix emerged as something new Whether or not that being used by anybody I don know It'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 west stuff enters my brain and just exits if i'm like i'm not i guess i'm not picking up remix i don't have room for it announce what it is going to be and it has nothing to do with react anymore and it is a sort of a full stack entirely javascript it still uses jsx it's all based on web standards we guessed it might be a shopify related product that is not the case. We guess that it might be a React server components based headless commerce thing. That was not the case at all. It has nothing to do with React anymore. Next one is React Native will have its time. React Native is kind of its own thing. It gets used a ton where it's needed, 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 nailed that one either. Yeah, I think it has. I think React Native continues to get better. Expo continues to get better. The more I see from React, one thing React Native, I think, just added CSS, better CSS support for a CSS grid or something. Oh, really? They have 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. So Lynx, for anyone listening, people are probably trying 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 launched their own version of React Native, which powers a lot of the big apps that they have in China. Yeah, Lynx is really cool. Interesting. I'll tell you that. Keep your eye on Lynx. Next one, Tanstack start and Tanstack will pop. Yes, we nailed this one. And Tanstack has got hype around it nonstop. All the Tanstack stuff. I've been seeing major things move to Tanstack. And I think they have a lot of... It's funny. I mean, the Tanstack ecosystem overall has, in my mind, legitimized quite a bit, but also grown quite a bit. They just launched Tanstack AI, which I think is going to be huge. Which we knew about early, by the way, folks. Yes. That's what Tanner told us after the episode we did. with him he told us what it was going to be that's pretty cool so just like a standardized ai sdk that can be used in javascript php and what else is it i think python which is really interesting they're going with like every language you can use this sdk and not just like tanzac 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 if we standardize on these ideas of responses and requests and the whole thing is going to be beautifully typed, I'm pretty excited about that. Yeah, I'm going to do some tutorial content on this Tansic 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 dope so that is component level data loading yes i also wrote 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 going to act like it's some brand new thing. But 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 Bunn, we had two predictions. We say Bunn will keep doing wild but loved non-standard features and they certainly continued that they they doubled down on their node compat this year but they also did a lot of non-standard features which was they essentially became a v competitor and they also implemented their own router so bun is an awesome tool but probably the big one we 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 going to do you know the the clear path forward is just like they're going to release a hosting environment that did not happen they got acquired by claude because bun is powering like claude code yeah i got 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 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 CloudCode is why they've acquired them and done it. So I don't know why they didn't just use CloudCode to spin up a BUN themselves. It's going to 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 do that. I don't think AI is able to write Zig because Zig just moved off of GitHub. So now their code isn't getting flurped 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 that's not Next.js is Vite. That's really where it's at. Or, I guess, bun. But, yeah. Rolldown. rolldown as a man rolldown was released but it's not the default yet in veet and rolldown it was the default in seven what was it not i don't know i'm so there's too many things wes uh yeah i'm pretty sure i was upgrading something the other day and i went through the breaking changes rolldown how to try rolldown you have to use veet npm rolldown veet latest to use rolldown the current docs say that veet is planning to integrate rolldown to use rolldown you have to use rolldown veet it is still considered experimental yeah you're right 7.0 they rolled it out and you have to opt in it's 8.0 which is in beta 1 as of today no where it will default to roll down v so probably in a couple months we'll see that be the default i've been using it on all of my projects and you have absolutely awesome yeah it's great yeah there's also one thing that we missed on this is v plus so v plus was what is it give give me the rundown oh man i'm gonna be honest i don't have a good answer for that question uh the whole idea behind v plus is not v just not being like the dev tool and bundler and builder that it is right now but it also is your your linter your formatter your test runner uh ui testing all basically absolutely everything and that's their their way to make money right 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 want to be able to pick my tools and maintain this sharded javascript landscape that i love i've grown to love so i i want to use v for everything and i already have only been using v for everything As long as it stays amazing, you know, and doesn't, and they've proven that to me over, what, eight versions where 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 is going to release a CMS. That did not happen. they said they were going to acquire or tightly partner with Statomic because at the time this was during all of the WordPress drama and everybody was like well I'm going to 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 they're still not a very good replacement for what WordPress is which is a huge ecosystem with lots of plugins 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 got like i think the biggest bytes are like things like squarespace and webflow and whatever they've taken bytes out of that but it's still so huge i don't think that the majority of WordPress users even know what anything regards to any of the drama even happened or was. So developers, sure. But like the users, no. Yeah, yeah, I agree. That's our predictions. Tallying it up. I think we got a healthy 87% accuracy. I think we did pretty good on that. What do you think, Scott? I think we got 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 2026 predictions yeah do you want to do a sick pick before we get out of here uh 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 woozu fan yeah it is this fan that goes it's a fan that goes on two axes 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 uh 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 kind of guy so i like a little fan on at all times in the when i'm especially when i'm yeah me too yeah yeah 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 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 uh that is how i hold it when i use it though yeah yeah no i've been using it actually uh well i've been using it to clean clean like in the kitchen or something like that i'm like oh man you see everything with this i actually got it for my skin uh believe it or not but like so wait what does it do it's a It's basically a really high-powered black light. And you see what? You see every little crumb ever. You can find all kinds of stuff. I lost a pill last night, and I was like, I knew it was here. And then I was like, oh, let me bust out the black light. I turn 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 black light and all of a sudden it's like right there. Now we can clean it better. I've been finding so many little use cases for this thing. And it's a nine bucks, this little black light flashlight. I've been using it to do all kinds of just like cleaning around the house or whatever. Or like when I'm cleaning the kitchen at night, I'm like, I'm not like a germaphobe or a clean freak or something. But I turned it on, I was just like, oh, there's all this crumbs here that I missed or whatever that you're just not seeing otherwise. So it's funny. I got it for a completely unrelated use case, and now I'm just using it for all kinds of stuff. Wow. Terrifying. Terrifying. It's not terrifying. Also, when you pointed at the glow-in-the-dark syntax sticker, it instantly turns neon green. That's pretty sick. 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 going to happen let us know where something that you thought was going to 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 going to happen in 2026 beauty peace Thank you.