967: What’s Going to Happen in Web Dev During 2026

Wes and Scott talk about their bold predictions for web development in 2026, from WebGPU-powered design and modern CSS breakthroughs to JavaScript standards, AI-driven tooling, security risks, the future of frameworks, workflows, and more! Show Notes 00:00 Welcome to Syntax! 00:49 WebGPU and 3D experiences will finally take off Lando Norris 01:30 Web design will make a comeback Raycast shaders.com 04:03 Light mode returns (yes, really) 07:06 Modern CSS standards are about to have a huge year CSS Wrapped Graffiti 13:15 Will the Temporal API finally ship everywhere in 2026? 14:18 The rise of the standard stack 16:18 Are we headed toward standardized RPC? 19:41 What’s next (and what’s not) for React 21:07 Why we’ll see more security failures in web dev 22:35 SvelteKit 3 lands in 2026 22:53 Where developer tooling is headed next Oxc Biome 26:44 More big acquisitions Anthropic Bun 28:02 2026: the year of durable compute 30:57 Frameworks will matter less as AI gets better 33:34 End-to-end AI workflows become the norm 36:04 Brought to you by Sentry.io 37:21 Personalized software for everyday people 39:11 MCP and MCP UI will pop 42:24 Developer skills will fall off 46:20 Crappy software will continue Hit us up on Socials! Syntax: X Instagram Tiktok LinkedIn Threads Wes: X Instagram Tiktok LinkedIn Threads Scott: X Instagram Tiktok LinkedIn Threads Randy: X Instagram YouTube Threads

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Scott Tolinski?

  Welcome to Syntax. We have our 2026 predictions for you today. What's gonna happen 2026? Last year, we had an incredible hit rate. I mean, we were well into the 80, 800%.

• 0:15Wes Bos?

  87.5% accuracy, yeah.

• 0:17Scott Tolinski?

  We guessed things that we didn't even know that were gonna happen. And today, Wes and I are standing up. In fact, Wes was like, "I'm standing up." And you know what I said? Me too. I'm standing up too.

• 0:30Wes Bos?

  New year, new me. I'm a stander now. I can't wait to try stand for a week and a half and then sit down for the rest of the year.

• 0:37Scott Tolinski?

  Yeah, maybe I should do an episode on the treadmill. Get walking. Let's get into it, though. Let's get our predictions cooking here. What do we predict for the year 2026 in the world of web development? So, number one here is WebGPU and 3D experiences pop. Man, let me tell you, that Lando Norris website made some waves. And that made some waves that won't be felt until 2026. But I think we're gonna really feel them with new experiences. Now, we saw a lot of really cool GPU-enhanced websites or websites that had a lot more shaders and graphical

• 1:20Scott Tolinski?

  elements and things like that. I think it's only gonna get crazier for 2026, for sure.

• 1:26Wes Bos?

  I think so too. Man, the cool landing page, and this, this kind of, I'm gonna loop this in with, with another one of my predictions, is design making a comeback because everybody's cranking out the same decent-looking website with AI.

• 1:43Scott Tolinski?

  Raycast ass homepage.

• 1:45Wes Bos?

  Yeah, in order for, and that's not a, that's not a slander at Raycast at all because Raycast like invented that whole UI. And it, it kinda sucks that, like, your entire...

• 1:58Wes Bos?

  Like aesthetic has become like a meme because it's so good and everybody copies it. But websites where you land on it and you feel like, wow, somebody has actually put some work into this. It makes you stop and look at it and experience and feel good about the product. Because there's so many generic ass websites where I can't even tell if I'm on like what website I'm on. Sometimes I'm on threads and Twitter and I don't even know what website I'm on. Cause they all look exactly the same. And, and part of that is good because it's like, you just know what to expect and you just want to use the application. But also like there's, there's some joy that's been lost in, in the website.

• 2:39Wes Bos?

  And I think the 3D, all these web GPU experiences are going to improve that. People are able to make new things. So Simon Lamarche is working on shaders.com. And it's just like really sick. It's a web app and like an NPM package that you can just like use shaders together and you can mask things and blend them. And there's a whole bunch of like defaults that you can, you can pick out. And I think it's the year, it's the year for 3Ds. It's the year that people are going to learn what shaders are and make cool ass looking websites.

• 3:14Scott Tolinski?

  We're going to see dithering all over the place. People are going to find out about dithering and they're going to dither.

• 3:20Wes Bos?

  And noise.

• 3:21Scott Tolinski?

  Yes. And I'm not just going to stop, drop.

• 3:23Wes Bos?

  Yeah.

• 3:24Scott Tolinski?

  I'm going to stop, drop, shut them down, roll them up, shop.

• 3:27Wes Bos?

  Stop, drop, grunge.

• 3:29Scott Tolinski?

  Stop, shop, drop, shut them down, open up shop. Sorry. Yes. I agree with you on that. Shaders, man, shaders are a crazy thing, but I also think AI will improve the burden on learning shaders.

• 3:42Wes Bos?

  I think that's why. Is that like in order to like write shaders right now or without AI? Like you have to be a mathematician, you know? You have to be an artist and a mathematician and. Basically, people want to be able to like not be mathematician and still make cool looking stuff. And I think the tools for that are coming this year.

• 4:03Scott Tolinski?

  Yeah. Yeah. Light mode makes a comeback. I gotta say, I've been using light mode. I would join.

• 4:11Wes Bos?

  Yeah, I told you.

• 4:13Scott Tolinski?

  There's evidence here. I, I, I joined the syntax team meeting the other day and I had light mode on. I had to turn it off because it was making my glasses look crazy. But yeah, that's the hard part about glasses where it's on camera, you know, gives you this look here. But you know, for me personally, I, I have been liking light mode during the day and dark mode in the evening. And I have the automatic thing set up and I have found light mode themes that I really like in my text editor for the first time in my entire life. I think the last time.

• 4:45Wes Bos?

  Your entire life? Not even like Dreamweaver notepad days?

• 4:49Scott Tolinski?

  I was gonna say, the last time I used a light mode theme for more than enough time to change it to a dark mode theme was Dreamweaver. Because the moment I went into TextMate, I would a dark theme. And then in Sublime Text, dark theme. And then dark theme ever since. So.

• 5:07Wes Bos?

  I'm just so sick of like crappy dark mode. It's like brown or washed out gray.

• 5:13Scott Tolinski?

  Notion is the poster child for that.

• 5:15Wes Bos?

  Doesn't work well, you know, I like, I love dark mode. I still dark mode when I'm coding. But most of my other apps default to light. And I like it.

• 5:25Scott Tolinski?

  I use a system setting for everything. And I was actually asking this on Twitter the other day was like, do people really want the little sun toggle in some personal person's website? Or do you just want it to be the system setting?

• 5:37Wes Bos?

  I do. Because I, I opt in on a per site basis. Like I use Twitter in dark mode. Yeah, it looks way better. And it's just, it's just about how it feels for that website. And like, like I said, there's so many crappy dark modes that have been half assed. So I, I want that little sun toggle. Don't, don't give me that all day.

• 5:58Scott Tolinski?

  Google Docs will ever get to dark mode? Like, come on. Like, what are they doing?

• 6:05Wes Bos?

  No, who's writing Google Docs in dark mode?

• 6:08Scott Tolinski?

  I would.

• 6:08Wes Bos?

  No chance.

• 6:10Scott Tolinski?

  Every time I go to the Google Docs, I'm like, come on.

• 6:13Wes Bos?

  No, no. No, no, never, never doing that. Well, maybe, we'll see. I don't think it will ever come. That's such a huge under. Yeah, to do all the icons, you know, but then they also have to, you also have to count for, people have made documents and slideshows, assuming the background is white.

• 6:33Scott Tolinski?

  Well, slide, I'm not talking about slideshows. I'm talking about like the canvas areas.

• 6:37Wes Bos?

  Yeah.

• 6:37Scott Tolinski?

  Stuff like that.

• 6:38Wes Bos?

  Like, but like a doc, somebody made a doc or somebody made a spreadsheet and the colors that they've chosen on that spreadsheet are based on it being a white background. So what do you, what do you do? Can you calculate a color that is similar? Yeah, you actually can. So maybe they can figure that out.

• 6:54Scott Tolinski?

  Come, that's what.

• 6:56Wes Bos?

  Please beep that out.

• 6:59Wes Bos?

  We don't swear on this podcast.

• 7:01Scott Tolinski?

  I know. I thought it would be funny with a beep. I thought that would be funny.

• 7:05Wes Bos?

  Yeah, that will be funny. Okay.

• 7:08Wes Bos?

  Next one we have here is CSS. Modern CSS standards are going to kick ass. Man, there is so much that has happened in the last, I don't know, maybe 14 and a half months. So much that is coming. Go to chrome.dev/csrap2025 and you will, you will feel like almost stressed out about how much stuff has been added to CSS. It's unbelievable, right? We've got customizable select CSS anchor, starting style, if statements, pop over, scroll animations and related to pop over, there's like eight different things, you know, like hover intents and hinting and style queries. And we've got CSS functions that are currently being added to the spec.

• 7:53Wes Bos?

  So I think that CSS is going to have a year this year and I'm here for it. I think a year from now, people are going to be like, wow, we can actually use this stuff. I think the Shad CN, all these UI frameworks are going to have major iterations or we're going to see some new version of these UI frameworks that are entirely built on these modern standards. You don't need to use some floating UI or a whole bunch of like external JavaScript libraries.

• 8:22Scott Tolinski?

  I'm just gonna take a minute to plug my, my thing. Graffiti is kind of built on that idea. Is that like, yeah, I'm not, this is not a slam, but that utility classes should have more utility. Right? Like, they shouldn't be just a DSL for CSS. It should do more stuff. So like, uh, you should be able to have a swiper or a, um, like I have an Instagram style reel that's just dot reel or, you know, their header is, applies a lot of styles. So it's almost like components based out of, uh, actual, actual, out of CSS without using any things other than, you know, the CSS modern features. And let me tell you, there's some cool stuff in this thing from the color system that's like alpha based.

• 9:06Scott Tolinski?

  Rather than like shades based, using relative color, easy to configure. But also, uh, a whole text based system that does, uh, like fluid text based on its, uh, container query or automatic contrast color with really, really cool relative color syntax. So I, whether or not it's graffiti, which I don't think it will be, but I think things like graffiti will, like you mentioned, will become more ubiquitous or systems that are already ubiquitous will lean less on JavaScript. Which, man, people in the comments of this YouTube channel, when I, or even my personal channel,

• 9:47Scott Tolinski?

  I did like a, a thing on Invokers about, oh, you can, you don't have to use JavaScript for this now. People are like, stop trying to kill JavaScript. Like JavaScript can do it fine. And it's like, brother, this is less code, it's easier and it's more performant. They're like, what do you want? You want to do it yourself? You want to have it be worse just for the sake of using JavaScript? Like, that's crazy.

• 10:13Wes Bos?

  I love JavaScript. And I agree with this. Yeah, like these things are primitives. The only reason we added JavaScript in the first place was because we couldn't use those things. Now we can. So, uh, yeah, let's do it. I, I think you gotta make this thing look like sick because like, there's like this like functionality point of view to it, you know?

• 10:33Scott Tolinski?

  Yeah.

• 10:33Wes Bos?

  But then there's also the reason people pick up Chat C N or, or Tail and UI or any of this is because it just looks good by default. And those defaults are becoming so ubiquitous. If there were something that looks different, not to say that you can't make Chat C N look like anything. Literally, you can make it look like anything you want. That's the whole point of it. Just people stick with the default. But if there's something that comes out with whatever that looks like next level, looks really cool.

• 11:00Scott Tolinski?

  Yeah.

• 11:00Wes Bos?

  I think that will, that will pop.

• 11:02Scott Tolinski?

  Yeah, man, there's so much cool CSS stuff. One thing I think that is underrated in the, the potential modern CSS landscape is scroll animations. I think because they aren't available in Firefox yet, or just that we haven't seen enough with scroll animations. Or I think there's a misconception that the thing that you can do with them is just

• 11:26Wes Bos?

  Parallax.

• 11:26Scott Tolinski?

  Scroll onto page, animate stuff in, or parallax or whatever. I think scroll animations have the most potential for like a major shakeup in terms of like interesting things that we can do in CSS.

• 11:41Wes Bos?

  Like, like interactions, you mean, right?

• 11:44Scott Tolinski?

  Yeah, all kinds of stuff. I always, I just saw an example using them in like form inputs or something. So, I mean, there's like a lot of like really interesting opportunities. But as this stuff gets better and better, I, I, I think we're going to see not just these features hit. We're going to see UI patterns like the swiper thing that I built, all of a sudden being like, oh, wait, we can do this. Oh, no, wait, we can do this. We, like, this always happens in CSS. A new feature comes in. And then people find out, okay, that's great. And then like three years later, they find out, oh, there's so much bigger possibilities with this. Now that we've really gotten comfortable with this API.

• 12:28Scott Tolinski?

  And now we can really explore some interesting concepts and ideas with it.

• 12:32Wes Bos?

  You know what is another cool thing I just saw on this Chrome CSS wrapped is scroll markers. So if you're doing a carousel, um, one of the things you still need JavaScript for is you gotta like, put all the buttons and then you gotta like, listen for like the scroll. What's the, what's the event in JavaScript? When the scroll snap snaps?

• 12:51Scott Tolinski?

  I don't even know.

• 12:52Wes Bos?

  Well, there's, there's an event for scroll snap snap and you listen for that. And then when that happens, you, you update your, like your dots as to which one is currently active. There's a new thing being added to CSS called scroll markers and, and scroll market marker target current, which will allow you to make buttons that

• 13:11Scott Tolinski?

  Yes.

• 13:12Wes Bos?

  Click next and whatever. CSS is so good. Speaking of CSS, let's talk about JavaScript.

• 13:18Scott Tolinski?

  Yeah. Speaking of CSS, yes, JavaScript, we said temporal will ship in every browser because every year it seems like that is, what is this? How many years have we predicted?

• 13:28Wes Bos?

  I think we've predicted it for probably four years now. And last year it hit, what it hit Chrome kinda, it hit, it's going to hit Chrome in 2026 January. So maybe not, but it's, it's not in Safari yet and it isn't.

• 13:44Scott Tolinski?

  This is a tap in Wes. This is a tap in because it is teed up on everything. I feel like, I feel like this will have to happen. Just a little

• 13:52Wes Bos?

  It's gotta, but yeah. But will there be like a gotcha? You know, it works, but Safari doesn't do seconds or something weird like that, you know?

• 14:02Scott Tolinski?

  Yeah. Safari thinks that there's a security reason. So we're going to completely limit the API. So that way

• 14:08Wes Bos?

  Safari does this one weird thing that just renders the whole thing unusable. But temporal will ship. This will be our year of temporal and I'm so stoked for it. Also, standard stack will gain popularity by framework author. So this is something I did a talk about this probably about a year ago. I've been talking about it a lot. There is a lot of new features in JavaScript. Navigation API, URL pattern. There's a new sanitizer API which will like, you know, we always have to reach for like DOM purify to like, you're taking user input. You want to put it in the DOM and you gotta strip out like a image on load, something like that. There's a new sanitizer API that's coming out. All of these new standard primitives are going to

• 14:55Wes Bos?

  Primitive in 2026. So you're already seeing this with like Hanno and Eliza and a lot of these standard server frameworks, but now like Remix as well. Now that they have become standards in JavaScript, I think a lot of these frameworks are going to pick them on up. So the standard stack is going to be gain popularity and that makes sharing code between multiple frameworks. It means it means moving code from one framework to another and makes all of that stuff super easy.

• 15:27Scott Tolinski?

  I think in general frameworks will continue their push to RPC territory where we've seen this with SvelteKit, we've seen it with everything, TanzStack query. People are moving to RPC, which feels like it's about time. Man, I used this kind of API in Meteor land. And like when I moved to have a separate API or like, like I was always just like, come on, this feels so much worse than how I was doing it in 2015. So I do think this is going to just continue to grow. And there's all kinds of things, whether that is like Convex or Superbase or these things that kind of obfuscate some of that and you're just calling functions and it handles it all

• 16:11Scott Tolinski?

  pocket based, those types of things.

• 16:13Wes Bos?

  Yeah.

• 16:13Scott Tolinski?

  Yeah, I think this is just going to continue to become more of the commonplace.

• 16:18Wes Bos?

  Do you think that we will see a, a push for standardized RPC? Because we have, like you have the Svelte version and then like Cloudflare has their own called, they rolled out called Cap'n Web and, and then there's obviously Rack Server Components.

• 16:34Scott Tolinski?

  No, I don't think we'll see a standardized version of this.

• 16:36Wes Bos?

  You don't think we'll...

• 16:37Scott Tolinski?

  I don't see that.

• 16:37Wes Bos?

  There are some, there's like JSON RPC, there are some specs out there for it, but I think as we move a lot of this stuff to RPC, probably not this year, but I think at some point there will be some sort of call of like, let's... Let's standardize these things and maybe we don't necessarily need to because of the fetch API. Like we have web request and web response and we have get, put, post. We have the body, you know, like those things are are fairly standard already. So maybe that's that's good enough.

• 17:11Scott Tolinski?

  I will say that I do think that standard schema Yes.

• 17:15Scott Tolinski?

  will become way more ubiquitous in regards to RPC. Even zero sync in their new mutations has moved to having a standard schema based the same way that it's felt. Remote functions are. So that way you can use archetypes odd, whatever you want. And to validate your data between, because that is really, as we, as we talked about in the remote functions episode. That is really such a killer way to validate your data between the client and server. So I, I, I would hope that that would, that aspect of it might be standardized.

• 17:53Wes Bos?

  I think so. If for those who don't know, standard schema is this idea that you take your, your library like, like ZOD or Valley Bot or there's several others, like 20 of them out there and that all of the tools that both ingest data, validate data, but also return data. They all speak this thing called standard schema and you can convert between them. And it's been so nice to be able to like, use whatever library I want, but just know that all of these tools work together. Whereas previously a tool would be like, we only use ZOD version three. And like, and that's it, you know, like, like the open AI API is like that right now. You can only use ZOD version three, but then you're starting to see a lot of these AI APIs being like, just pass us

• 18:38Wes Bos?

  standard schema and, and it will, it will work between all of them. Even like, what did I use the other day? Chrome's local AI API uses JSON schema. If you want to be able to specify the return value as like a like a structured content rather than just text, you have to use JSON schema. And I was like, come on. Use like, let me use ZOD or whatever I want and then just pass you the standard schema version of that. Don't pigeonhole me into using your weird version of this.

• 19:12Scott Tolinski?

  I feel like these libraries, whether that is archetype or whatever, are all just going to move to support both. And it's like, you can just use whatever you want, right?

• 19:22Wes Bos?

  But that's, that's the whole point of standard schema is that.

• 19:24Scott Tolinski?

  I know.

• 19:25Wes Bos?

  You can use whatever you want. You can specify it in whatever way you want. And then when you pass it to the tool, it doesn't matter because it.

• 19:33Scott Tolinski?

  Doesn't matter.

• 19:34Wes Bos?

  You can convert ZOD or archetype or whatever to standard schema. And then the thing will just ingest the standard version.

• 19:41Scott Tolinski?

  Yeah. What do you think about, what do you think is going to happen in React land this year?

• 19:46Wes Bos?

  Oh, that's a good question. I don't think much. I think like the pace at which we move with a lot of these frameworks, especially that so much more of the attention has been off of features and on to how do you do this with AI? I don't know. Like there's obviously, we had the huge React security hole. That's certainly going to leave a bad taste in a lot of React users' mouth, but I don't know. I don't think much. I think it's still going to be the number one, but I'm rooting for a lot of these other, other players in the space as well, you know, whether that's felt or remix or whatever.

• 20:24Scott Tolinski?

  Yeah. I think there will be another security issue. And I only think this because I was talking to CJ and he, he kind of opened my mind up to the idea that when you're inventing such a unique transport, that the possibility for something like this is opened up. So I do think there probably might be something else.

• 20:44Wes Bos?

  Yeah.

• 20:45Scott Tolinski?

  Based on nothing.

• 20:45Wes Bos?

  Around like server components initially, like you, React server components invented.

• 20:51Wes Bos?

  The whole transport end to end, right? And a lot of people were vocal, even Rich Harris, creator of Svelte, was like, we will regret reinventing RPC.

• 21:05Wes Bos?

  And here we are with a major security hole in it. Well, maybe we'll parlay that into one of my other one is we will see a lot more security mishaps in web development in general. I think that is because of several things. AI code slop, obviously, that's a big one. Understaffed tech teams, that's, I think that's been a lot of the reason we've seen so much downtime lately. And as well as like so many of these companies are just shipping at insane speeds right now. Because everybody's trying, it's a race to, to be the winner of all of this AI stack. And, and because of the insane speeds that they're shipping at, things obviously get missed and whatnot. So that's, that's a bit of a trade off or a huge trade off right there. And I think a lot of that stuff is starting to catch up with us, you know?

• 21:50Wes Bos?

  We've seen major downtimes in a lot of the, the cloud providers. And I think that is going to continue through the web development industry, unfortunately.

• 21:59Scott Tolinski?

  I think we will find at least one, at least one, I think there's going to be more. At least one major story that comes out about an app that people are really using. That's maybe not mainstream, but taken off. That once the security issue is found, the entire app will have found to be vibe coded garbage. Held together with duct tape and then we'll all laugh about it and et cetera. It'll be all over online. Like, I can't believe they did this. They, they called this database query from the client for some reason, you know? Uh, I think Svelte, for Svelte, we didn't talk about that. I think Svelte Kit 3 will be released. I don't think that's a difficult prediction to make.

• 22:45Scott Tolinski?

  And we'll include all of their async and remote data stuff all buttoned up and fully launched in, in a stable.

• 22:53Wes Bos?

  Let's talk about tooling as well. Like around all these web development tools. A couple of years ago, we predicted people will fall out of love with YesLint. That's happened a lot slower than I imagine. And it just says something about. The ecosystem that things are built around, it's the same thing with WordPress is that as as much as you don't like some of these tools based on like the UX and the DX of working with them, they're just so big that it is very hard to to move forward with. But we're starting to see some pickup in in people. I've moved a couple of projects to biome, which I've been been really enjoying. Oxlin is another one that's built into VIT plus. We'll see. So VIT plus is their

• 23:36Wes Bos?

  VIT's like paid initiative, I guess, where it's like an all in one tool that handles all of your linting and formatting and it's supposed to replace everything. I'm curious to see if that gets pickup or not.

• 23:50Scott Tolinski?

  I think, bold prediction, I think by this time next year, biome is going to get snow packed by Oxland. I think

• 24:00Wes Bos?

  Interesting. What do you mean? And what is snow pack? Explain what that is.

• 24:04Scott Tolinski?

  Snow pack was kind of like the first ESM bundler that was came out and then VIT came out and was better. And VIT kind of instantly took over and people stopped talking about snow pack.

• 24:19Wes Bos?

  Yeah.

• 24:19Scott Tolinski?

  I think biome and their approach does not allow for what I have heard, I'm not an expert on this, does not allow for plugins to be compatible in the same way that they were with prettier. This is me speaking completely uneducatedly on this and I should probably do some research. But I think because of how Oxland is built and the people behind it, I think Oxland will become the default non prettier formater and the default non prettier winter.

• 24:58Wes Bos?

  Yes. And the reason, I think the reason you're saying that is because biome plugins are written in something called grit Q L, which is essentially like you have to write some language that you are probably not familiar with in order to make these plugins and that makes the ecosystem. By default, not as strong. And that's why ESLint is so big is like any idiot can go and make like myself included can go write a JavaScript plugin for this. Whereas Oxlent is written in Rust, but allows you to write plugins in JavaScript and they are working towards implementing the entire spec of ESLint so that you will be able to just take your like I have some weird ESLint plugins in

• 25:42Wes Bos?

  some of my projects where it's just like, I'm not, I don't want to fuss with it. I hate I my most hated thing is working on the tooling and trying to get it to work. And when I hit save in my editor and it just formats and lends properly, not touching it until until the next time that it breaks. And then I'm like, oh, damn it. I gotta spend some time figuring out why this isn't working.

• 26:04Scott Tolinski?

  Yeah. Yeah. Do we think that anything will replace StyleLent? Or are people just going to continue to ignore StyleLent? Because StyleLent is fantastic. Hmm. I don't know. StyleLent has such a big ecosystem plugins. And even if these things say that they can lint CSS, I feel like that would be tough to replace everything that StyleLent is doing. I would love to see that, though. I would love to have everything lent in one tool. But I do think that Oxlent is going to be it and not biome.

• 26:38Wes Bos?

  Interesting. All right. That's we'll put money on that.

• 26:41Scott Tolinski?

  Planting the flag.

• 26:42Wes Bos?

  I'll see you on New Year.

• 26:44Scott Tolinski?

  Yes. Yeah. The next one is we saw Anthropic purchase Bun. I think we will see more acquisitions like that happen. I don't know who or what, but I do think more web technology properties, whether that is things like Bun or otherwise, will get scooped up by who knows who.

• 27:07Wes Bos?

  Yeah. Like a who needs Dino right now? You're thinking about?

• 27:11Scott Tolinski?

  Yeah.

• 27:13Wes Bos?

  Looking at all this.

• 27:14Scott Tolinski?

  Does anybody need Dino?

• 27:15Wes Bos?

  Who needs Tailwind? Someone should buy, might buy Tailwind. Like OpenAI or Anthropic has to buy Tailwind simply as a thank you for all of the vibe coding they crap out in the last two years, you know?

• 27:28Scott Tolinski?

  Oh, yeah, I did. It is interesting and I do wonder about, about some of these. I feel like Cloudflare is poised to buy some stuff, you know? Cloudflare bought, did they, they didn't buy Hanno or they, they?

• 27:41Wes Bos?

  Well, no, they employ the, the author behind it. They Aqua hired a couple of people from Vercel. Yeah, yeah.

• 27:49Scott Tolinski?

  Yeah.

• 27:50Wes Bos?

  There, there will definitely be some

• 27:52Scott Tolinski?

  Partykit.

• 27:53Wes Bos?

  aggressive moves. As, as much as these AI things are good at writing code, people that know what's up and how to write code are still in heavy demand. Here's one of my thoughts is 2026 is going to be the year of durable compute. And, and what I mean by that is this whole idea of plausible, waitable workflows that you can, can pause and resume. So probably the big ones are like temporal and ingest where you could simply have a function and you can await inside of it, or you have like a step and you can await for like four days and then four days later you can, you can pick up and keep going. Or you can try something every, every hour and then loop over five times.

• 28:34Wes Bos?

  You know, a lot of the stuff that we previously would use queues for, it makes a lot more sense to just use a function that reads top to bottom and can sort of go to sleep. And then you can wake it back up and you have all of your state and all your variables and everything that you were used to in the past. So ingest and temporal were two of the like sort of forefront in this area. Cloudflare workers has their workflows. Vercel launched their own workflow.dev, which I previously said was next JS only. That is not the case. It literally works with, with anything. In fact, they're working on a Cloudflare adapter, which I think is, is the move if you want people to trust it.

• 29:16Scott Tolinski?

  Yeah.

• 29:17Wes Bos?

  Armin, he used to work at Sentry. He's writing one called Absurd, which is based on Postgres. What else is there? Trigger.dev. There's so many people going to this and partially that's because of agents. And partially because like this is just a nice way to write code. You know, you have to send an email. You want to wait for that email to send and come back, or you want to try it several times over. The durable function is the, is going to be big in this, the year of 2026.

• 29:43Scott Tolinski?

  Yeah, I think that's interesting. I am curious. And by the way, just to be clear, you're talking about Temporal, the company.

• 29:51Wes Bos?

  Temporal.io, which is, has nothing to do with the Temporal API we just talked about. Yeah. This is a, a company that provides the, the ability to pause code.

• 30:02Scott Tolinski?

  Yes, the ability to pause code. Yeah, I think, I think you're right on this. I have not used any of these tools before. So may, maybe, uh, maybe I'll pick them up. Maybe it'll hit the, the mainstream.

• 30:12Wes Bos?

  Yeah. Any, any, like I even think about our like transcription service right now. You know, if you have a function that runs for, sometimes you have a function that runs for like 60 seconds because you're waiting for some heavy thing and you don't want it to timeout. Anytime you have like what would normally be a long running function or a function that needs to retry every so often, or a function that needs to, that could possibly throw and you want to retry in, in a couple of minutes. A durable workflow is probably the move there.

• 30:41Scott Tolinski?

  I look forward to, uh, hearing more about this in the, in the future, Wes.

• 30:45Wes Bos?

  Yes.

• 30:46Scott Tolinski?

  Possibly from you. Yes.

• 30:47Wes Bos?

  I'm working on a show on it right now. I have a card. That's why I was able to rattle off all of the names of those because I've been, been trying them all out.

• 30:56Scott Tolinski?

  I love it. Uh, let's talk about AI stuff. I think the AI, here's what I think. I think the framework will matter less because the AI tools get better.

• 31:08Wes Bos?

  She said this last year.

• 31:09Scott Tolinski?

  And that the AI tools will simply just pick whatever is going to work the best. Now, I, I do think that and, and I think I've personally seen it, where I'm, I'm not having to write a framework all the time, or I'm not having to, I'm not having to worry about the quality of the code based on which framework I'm picking. Now there's a lot of alpha libraries and stuff I'm picking. Picking up that, obviously I still do, but I think the tools will just can continue as we saw over the past year to get better to the point where I can say, write me a whatever app with this version and it's just gonna get it right instead of making up as much syntax.

• 31:48Scott Tolinski?

  Cause the amount of improvement I saw over the course of this year was massive in that territory. So I think it will continue that path.

• 31:56Wes Bos?

  I'm curious. So I'm pretty bullish on not like remix itself, but the idea of of remix where it is so much based on standard APIs.

• 32:07Scott Tolinski?

  Yeah.

• 32:07Wes Bos?

  That it just knows about fetch, web request, web response, URL search params, routing parameter matching. It knows about all of these things because they're just, they're primitives built into the language. And then if you can build your framework on top of that, it needs to know a lot less about the framework. And it needs to know a lot less about like the oddities of how things work. So I'm hoping that that means that you're able to just give it these brand new frameworks and it can do a good job at writing that code. I don't know. We said this last year and it's still kicking out React and Tailwind 3 for me unless you explicitly go and,

• 32:50Wes Bos?

  and do a whole bunch of work and it's not good at writing Svelte 5, you know?

• 32:54Scott Tolinski?

  Yeah. It's good at Svelte 5. I'll tell you that now. Opus 4.5. Good at Svelte 5.

• 32:58Wes Bos?

  It's good at Svelte 5. Dude, so you provided like the docs with it and it knows how to reason about that.

• 33:06Scott Tolinski?

  It's just good at it even without the docs now. This Svelte team made like an API benchmarks. That was like determining what percentage of Svelte features that it got correct.

• 33:16Wes Bos?

  Yeah.

• 33:16Scott Tolinski?

  And it reached 100% with Opus 4.5. So, yeah.

• 33:20Wes Bos?

  That's what we need because so much of this is just like, I feel like it's worse or I feel like it's better and.

• 33:24Scott Tolinski?

  No, the team has benchmarks and they've been updating it anytime anything new came out. Benchmarks. We will see more benchmarks. We will be continuing to benchmark AI stuff.

• 33:34Wes Bos?

  I think end to end will become much more popular. I called this tip to tail on a previous episode, meaning that having AI involved at every step, not just like you just boop, boop, boop, and then it goes off and rips the whole website for you, but having AI involved in your planning, your writing code, your iterating, testing, it's getting really good with Playwright and integrated browsers into these agents where it will can actually scroll the page, take screenshots, click things, drag and drop. I think having AI in the loop for every single part, rather than just writing the code and you having to copy paste an error from your terminal into the chat. I think that those improvements will get much better.

• 34:18Wes Bos?

  And I think that will significantly improve the speed at which we write this software and quality.

• 34:24Scott Tolinski?

  I totally agree. Now, I think there's even going to be new tools and new ways of working with AI that we haven't seen before. I feel like all the time, stuff comes out that improves the tooling around the AI and therefore the output of the AI ends up becoming better. Have you heard of beads?

• 34:45Wes Bos?

  No.

• 34:46Scott Tolinski?

  Yeah, beads is something that I just found out about.

• 34:50Wes Bos?

  A memory upgrade for your coding agent? Is that it?

• 34:54Scott Tolinski?

  Yeah. So what it does, if I'm, if I'm understanding this correctly, because I've only just started to use it, is it is basically taking that to do list that many of these agents are keeping in context and moving it into JSON and SQLite. And you're able to commit it to your repo and stuff. So therefore, that to do list of what the agent is using at any given point in time is less ethereal and more concrete and shareable across your team. And it is kind of like these systems that many of us are just kind of haphazardly doing in Markdown right now.

• 35:34Wes Bos?

  Yeah.

• 35:35Scott Tolinski?

  You know, like this product, this status of this, open, close, whatever, blocked by, it basically allows you to have a Git structured version of that type of workflow.

• 35:47Wes Bos?

  Your whole process through something.

• 35:49Scott Tolinski?

  Yes, I've just installed this on a project of mine, so I'm yet to see if this thing will really solve my problems, but I've heard a lot of really good things about it and it's got 4.3 thousand stars on GitHub, seems to be a rising star in this space, but I think things like that, as well as tools like Sentry, at S E N T R Y dot I O, Sentry.

• 36:11Wes Bos?

  Yeah.

• 36:11Scott Tolinski?

  Sentry has some really great AI tools and they fit perfectly into your normal workflow, where like, even in development, you can use the Spotlight MCP to be like, "Ay, what is this issue that's going on? And how can I, how can I solve this?" But in your, your production code too, like, I, I've found that their, their Seer platform, the ability to take any given issue and find the root cause of it, has been really nice. I think we're gonna see like, a lot more things like this, because when you use these tools in practice, they, even if it's just a little button in your Sentry that says,

• 36:54Scott Tolinski?

  "What's the root cause of this?" Like, I was on Netlify today, and I had a deploy that failed. And you know with deploys that fail, sometimes it's, "Who knows why the hell this deploy failed?" There you have an AI button that says, "Why did this deploy fail?" And you click it, and it gives you an AI summary of why it failed. I didn't have to go pawing through the logs and make a guess. And I think we're gonna see little things like that just pop up and continue to get better.

• 37:21Wes Bos?

  Next one you have here is, "Regular people will realize that they can make personalized software." And I know people hate this.

• 37:28Scott Tolinski?

  I don't know if regular people will, but.

• 37:29Wes Bos?

  But.

• 37:30Scott Tolinski?

  Developers will.

• 37:32Wes Bos?

  Developers already know this, that like, I have made so many little apps, my Christmas lights, I needed to do the spacing of the window frames for my Christmas lights, and I just had to input the width and height, and it told me exactly how far along, and it kicked out a 3D printed jig that I could then print and, and use to drill holes exactly spaced correctly in it. And like, that piece of software, I would never find anywhere, and it was so helpful to myself. And I think regular people also have the need for that, and regular people already do this with spreadsheets, you know, like, you should see some of the crazy stuff people do in spreadsheets to simply, and spreadsheets are an awful way.

• 38:13Wes Bos?

  To like to do stuff, right? You had to nest all these things in one single line. Imagine the people that have the smarts of making complex spreadsheets can now just move that over to like vibe coding and actually build software that's useful for them. I don't know what the avenue for that will be, right? Like you have like Bolton V zero, which is kinda something like that already, but there's gotta be one more step where like I'm still not talking to parents at school pickup or or whatever, you know, they're maybe using chat GPT for these things, but no, they're not making software yet.

• 38:48Scott Tolinski?

  Yeah, they're not making software. Maybe that will happen. Yeah, I feel like I feel like that's too soon, but who knows? You know, maybe we'll have the one thing that makes it really easy. You do see a number of these apps pop up.

• 38:58Wes Bos?

  It's gotta roll out in like, like Microsoft, whatever, you know, Microsoft App Maker. It's gotta come with like Microsoft Teams and like, oh, I'll just make a quick little app, you know, it's gotta be integrated with that. Well, I don't know what that will be, but I do know that MCP and MCP UI will will pop this year, you know? We've seen major improvements in MCP and we're starting to see quite a bit of work on the MCP UI front. Meaning that like we saw OpenAI release their half ass version of being able to integrate custom UIs into chat GPT. But that spec is severely limited.

• 39:38Wes Bos?

  And if we want to be able to just embed our own custom UIs into any chat or any AI agent, MCP UI is is the move for that. And I think that we're gonna, we're gonna see a standard. I think the MCP UI standard is going to be accepted by the MCP spec. And I think that we'll start to see it implement in a lot of these chat agents. I'm curious to see what OpenAI does around this because I know that they want to have the upper hand of control. They they might have like an Apple moment where they're like, no, we're gonna implement it, but in our own crazy way. And unfortunately, they can do that because they are the big one, the big dog in this space.

• 40:20Scott Tolinski?

  I'm woefully under researched on MCP UI, not something I have really looked at at all yet.

• 40:26Wes Bos?

  Awesome. See, like, imagine, imagine you are in chat and you ask it to like, I don't know, uh, show me the latest pizza places or display the latest syntax episode. And imagine that we could just return, instead of just returning data from our MCP, you could just return like an iframe with HTML and CSS inside of it. I don't know that this is the end. A lot of people are saying like, this is the replacement for the web.

• 40:53Scott Tolinski?

  For the browser.

• 40:54Wes Bos?

  You know, this is, the browser is dead and we're simply just returning widgets from, from that. And partially, I think like WeChat in China, a lot of commerce, a lot of ordering is all done on WeChat widgets. And it, it works over there. So I think there will, there will be a lot in which I don't think it's going to replace the entire web. I think it's a crappy experience for a lot of stuff, but I think for some commerce, ordering food, doing research, um, simply just returning UIs from, uh, from a chat instead of just text is, is going to be great.

• 41:35Scott Tolinski?

  Yeah, there are, there is a lot of times when like even just wanting to know a quick answer. I mean, AI has gotten good to the point where many times you're not searching the web for things, you know?

• 41:48Wes Bos?

  Yeah.

• 41:49Scott Tolinski?

  Asking quick questions, stuff that's easily verifiable or reasonable.

• 41:53Wes Bos?

  Yeah, they're already kind of doing that. Like they're returning graphs sometimes, you know? And, but like, imagine that you could just open that up to the entire web and say, you can return, if you install your Expedia or your Airbnb or you install the syntax MCP server, it should also be able to return like a custom UI, whether that's just nicely formatted data or whether that's actual like buttons that you can click and edit your data inside of that.

• 42:22Scott Tolinski?

  Yeah. That's reasonable. Yes. Next one. Devs will fall off. I think this is something that anybody who as a developer who gets deep down the AI rabbit hole doesn't realize it's happening until they're like, oh, I'm not using really my brain. I'm feeling productive. Things are coming out. Sometimes the code is really bad. If I look at the code for too long, it hurts my eyes. But you're not really, you're not realizing that you're not gaining the same types of skills if you're just having it spit out. It is so funny. We talked about this a number of times on this show, but when you, when you talk or when you read comments

• 43:04Scott Tolinski?

  in YouTube from people who are using a lot of AI, you can tell exactly where they are as a developer or in their AI journey, but based on like the types of things they're saying. And you always want, oh, you'll figure it out eventually. You'll, you'll, you'll get to why we're saying this. And people like just assume that you're either a hater or you're a maximalist or whatever on all of these things based on what you're saying.

• 43:27Wes Bos?

  People are obnoxious on the entire arc of I hate AI to it's going to replace absolutely everything. And if you're an absolutist anywhere on that arc, you're obnoxious because it's, we haven't figured it all out yet. It's very bad at some things. It's also very good at some things. And it's a lot of it. It depends. But you're getting stupid by being an absolutist along that arc.

• 43:50Scott Tolinski?

  Yeah, you are. There's, there's a comment on one of our videos about, I was saying to only use AI to help you learn rather than to have it do it all for you when you're trying to learn a new topic. That seems reasonable, right? If you have the AI do it for you, you're not going to learn it. That's if the goal is to learn the thing. And somebody responded with, this is a boomer take. AI is a revolution. Boil frog boil, as in I'm a frog boiling in a pot of water because the heat is turning up slowly.

• 44:23Wes Bos?

  That's hilarious because you're like the most anti-AI guy I know.

• 44:30Scott Tolinski?

  It's just like it's it's crazy town out there, but I do think that where you are in your journey will dictate and you'll eventually be like, ah, shoot, I'm maybe not picking up the same stuff. And maybe we'll have to learn less and we'll just be kind of dumber because of it, but we'll still be effective. Who knows?

• 44:46Wes Bos?

  Some of this stuff will literally not matter that we don't know how to do it anymore. I I agree with those those comments as well. Some of this stuff is like, oh, nobody can can write a nested selector or nobody can do X, Y, and Z anymore. And like, maybe that doesn't matter anymore. However, I I do know that you have to continue your skill set. And AI is very good at helping you push into new skill sets and and learning them. Because if you're just say, well, it does all I want now. I'm just gonna push the button that and I think you're gonna be the one that's gonna fall off with this. Because you're not going to continue that arc into becoming smarter and learning new things.

• 45:29Scott Tolinski?

  Yeah.

• 45:29Wes Bos?

  So we're keep learning things, folks. You know what I learned the other day, Scott?

• 45:34Scott Tolinski?

  What did you learn?

• 45:36Wes Bos?

  RFID and and and why am I so stupid?

• 45:43Wes Bos?

  RFID and NFC. This is like a little Proxmark III. And like, I spent a day diving deep into how NFC works, the different types of it, how how these different cards work, which which ones are writable? How bits and bytes are stored on an NFC card, you know, I've got, I've got like a whole bunch of them, got one right here and these little tags. And like, I feel a lot smarter after having known all that works and I've got a lot more lower level than I do with a lot, lot of my skills. And you just gotta keep doing stuff like that.

• 46:18Scott Tolinski?

  Yeah, keep doing stuff. Yeah. Yeah, I think the insidification will continue, as we've seen all kinds of avenues here, software feels like it's falling apart. Products feel like they're falling apart in general. And unless something big happens, like I think we're we're due for another year of things feeling like they're getting worse for some reason. As everybody's just trying to get that bag, baby. But that's, that's, yeah.

• 46:44Wes Bos?

  Because everyone's trying to get that bag right now.

• 46:47Scott Tolinski?

  Yeah. It's tough. It's tough. I, I, I do see, and this is kinda crazy. I do see a world where AI eventually helps with the entidification.

• 46:57Wes Bos?

  Yeah, like you mean like, like makes things better, you know, cause like we're talking about all of these security things and, and whatnot, um, and all these bad things, but like the hilarious thing is that, that these tools can also significantly help you find

• 47:11Scott Tolinski?

  Totally.

• 47:11Wes Bos?

  slow code and help you find security issues and vulnerable. So that's what you're talking about?

• 47:16Scott Tolinski?

  Yeah, help you fix problems, help you write more, uh, more defensive code from the start instead of just something that works.

• 47:23Wes Bos?

  Yeah.

• 47:24Scott Tolinski?

  Yeah.

• 47:25Wes Bos?

  I agree.

• 47:25Scott Tolinski?

  I agree.

• 47:26Wes Bos?

  Yeah.

• 47:26Scott Tolinski?

  Yeah.

• 47:27Wes Bos?

  Beautiful. Let us know. That's our 2026 predictions. Let us know down below what your predictions are. What you think we are wrong or right about something as well as if we didn't hit any of these things. Cause maybe what we'll do next year is we'll come back to the comments of this video and say, what did, what did the folks in the comments, what is Mr. Paperclip who wrote something about it? What is it? Where are we at right now with this type of thing?

• 47:51Scott Tolinski?

  Yeah, Mr. Paperclip. Yes.

• 47:55Wes Bos?

  Cool. All right. Thanks everybody for tuning in. We will catch you later. Peace.