Transcript
00:00:00Last week, we had SvelteKit 3 and Solid 2 being released as release candidates.
00:00:08So not fully stable yet, but almost.
00:00:10And a couple of years ago, this would have been an interesting thing.
00:00:15You would have seen a bunch of videos about it, including by me, of course, maybe even a course.
00:00:19And the question nowadays is, does that still matter?
00:00:24Do new frameworks still matter at all or new framework versions as we have it here?
00:00:30Or does it not matter because with AI, it's all about React anyways, right?
00:00:36And why would you even care?
00:00:38I mean, if AI writes so much code, maybe all your code, why would you care about the underlying framework or library?
00:00:46And I'd say you should care quite a bit and new frameworks and libraries do matter.
00:00:52But, of course, gaining traction, I believe, is much, much harder than it used to be.
00:00:59And it was never particularly easy necessarily, but it may be worse now.
00:01:05But let's dig in.
00:01:06Now, I will not take a closer look at what exactly these frameworks or versions are about.
00:01:13I will link the official announcement blog posts, though, so you can dive into that.
00:01:17But I will say that both are pretty interesting.
00:01:21And indeed, I plan on using SvelteKit 3 specifically in some of my new projects, internal projects for now, to see how well it works with AI.
00:01:33Because that is important.
00:01:36One reason against these new frameworks or against the importance of new frameworks that speaks against it would be that AI has a favorite stack.
00:01:49And that, of course, is React and so on.
00:01:52So React and Next.js maybe, etc.
00:01:55And that is true.
00:01:56And that is true.
00:01:56That is definitely the case.
00:01:58But AI is pretty good at working with new, unknown things, too, as it turns out, if you as the developer give it the right context.
00:02:09And that is not to be underestimated.
00:02:12Of course, if you're just vibing away and you're not steering AI in any way and you don't care about any decisions it makes.
00:02:21So if you're vibe coding, which can be fine for certain one-off projects or throwaway software.
00:02:28But if you're doing that, then, of course, yeah, it'll not know what SvelteKit version 3 is necessarily.
00:02:35And you wouldn't care about it.
00:02:37You wouldn't even tell AI to use that version.
00:02:39So obviously, yeah, new frameworks or libraries in general don't have a place in that world.
00:02:46But it's not all about vibe coding, obviously.
00:02:48It has its place.
00:02:50And I don't know which role vibe coding will play in a few years and what all these tech choices will play, if that will still matter in a few years.
00:03:00But right now, you can definitely get a lot out of actually steering AI.
00:03:07That's kind of an understatement.
00:03:08And you should be steering AI.
00:03:11Our role as developers simply is changing.
00:03:15We still should be controlling what we build and how we build it.
00:03:19And that we don't write the code anymore necessarily does not mean that we should not care about the code, the architecture, the decisions we make.
00:03:29That is as important as ever or maybe more important than ever before.
00:03:34I think the reasons for picking a framework over another are different now.
00:03:42You know, back in 2019 and so on, one major reason for picking a certain framework amongst others, of course, was the developer experience.
00:03:54Do you have a good time working with it as a developer?
00:03:58So obviously that mattered, you would be spending a lot of time writing code in a certain framework or with a certain library.
00:04:04So obviously you want to have a good time.
00:04:06Now, of course, you don't always have that choice.
00:04:09You may be employed somewhere where you are simply forced to use a certain framework or library.
00:04:15But as a developer, depending on the company where you're working, and especially, of course, if you're self-employed, you can have and you could have, most importantly, some leverage and some impact on how decisions are made and which technologies are used.
00:04:29So definitely it mattered back then if a framework has good developer ergonomics, if it has a good or if it offers a good developer experience.
00:04:41That was not the only thing, of course, back then already performance mattered, bundle size mattered.
00:04:46Then there are other things that may matter.
00:04:49For example, how easy or hard is it to do server-side rendering with a given framework if we're talking about JavaScript front-end frameworks and so on.
00:04:58So lots of important aspects, but the developer experience aspects are less important now.
00:05:05I would say developer experience doesn't really matter anymore.
00:05:10Now, what will matter is the AI experience, I don't know what we call that, is it AX?
00:05:16That, of course, matters a lot.
00:05:19And things like the performance and the bundle size and all that stuff also still matters, of course.
00:05:27And I would argue you could say with AI now, because AI is so fast at spitting out code, obviously, it may actually matter less which framework you use.
00:05:43But then switching is easier than it used to be, and you can therefore really see and take a look at which aspects matter for the app, the project you're working on, what's most important for you, and then find the best framework or library for that and just use that.
00:06:00And you don't have to care about developer experience or anything like that.
00:06:03And I would say you don't even have to care as much about the ecosystem anymore.
00:06:07The ecosystem, of course, still matters.
00:06:09It is a factor, for example, if you're working with React, you're obviously buying into a huge ecosystem.
00:06:17You have ShadCN, you have a bunch of other libraries.
00:06:20But one aspect that matters less, of course, unfortunate for me, but that matters less is the amount of tutorials you have on a technology.
00:06:29Because, again, you may not be writing the code yourself anymore.
00:06:33So that's one aspect that's not as important anymore.
00:06:35And also, the other part of the ecosystem, the libraries and so on, yeah, sure, it's nice to have that.
00:06:41But AI is pretty good at just building whatever you want it to build.
00:06:46And if you want your own ShadCN for Svelte, and I know there already is a ShadCN for Svelte.
00:06:54But if you wanted to build your own version of a library that exists for React, but not for your favorite framework, you can just ask AI to do it.
00:07:02You can just ask AI to fork some popular React library and port it to your favorite framework.
00:07:11Now, that will not be worth the effort for every little side demo project you're building, but for bigger projects, it very well may be.
00:07:19You can do stuff like that.
00:07:21So, actually, I would say the ecosystem still matters, but is less important than it used to be.
00:07:28Now, what matters, though, and I mentioned this before, is the AI experience, you could say.
00:07:33And what do I mean with that?
00:07:36There are probably different things you could mean here.
00:07:39What I mean is how easy is it for AI to learn or understand the thing, the framework, the library you want it to work with?
00:07:47And how much does this framework force AI to write good code?
00:07:54And I'll give you an example.
00:07:55There is eFact.
00:07:57Now, eFact is not a JavaScript front-end framework, but it nonetheless is a JavaScript or a TypeScript library designed to help developers easily create complex, synchronous and asynchronous programs.
00:08:08You can, for example, use it on the backend.
00:08:10It has a bunch of stuff included.
00:08:12It's really amazing.
00:08:13And one thing that makes eFact amazing and probably one of the reasons why it is popular in a certain niche right now, at least, is the fact that it comes with a lot of rules that are actively enforced and that will yell at you if you're, or the AI, are writing the wrong code.
00:08:35If you're violating certain, well, rules of the fact library, because it's a TypeScript library, it heavily embraces TypeScript, it does a lot of type checking and a lot of other stuff that can really force you to follow a certain set of rules to write your programs in a certain way.
00:08:54And of course, when working with AI, anything that enforces rules and that yells at AI, if it invents stuff or if it tries to break out of those rules, anything like that can be good.
00:09:09And that's one aspect of a good AI experience, I would say.
00:09:12Working with a library or a framework that enforces certain rules and that basically minimizes the amount of mistakes you can make or the amount of variance you can have in your code, if you want to put it like that.
00:09:28So that's one aspect.
00:09:29The other aspect is how easy it is for AI to learn something.
00:09:32And here, it's, of course, all about the docs, because I mentioned earlier that with AI, you have something that is pretty good at working with new unknown things, too.
00:09:44Sure, it has all that baked in knowledge about React, but to some degree, you could even argue that this is a bad thing because it learned a lot of bad patterns from all these public repositories.
00:09:56But you just have to point AI at the right docs pages or give it the right skill, and it can do pretty, pretty well at working with a certain technology.
00:10:07I'm doing that all the time.
00:10:09I'm using AI with new versions of libraries or frameworks.
00:10:13I'm using it with Effect, for example, which is not new, but version 4 is new.
00:10:18I'm using it with something called Alchemy, which is an infrastructure as cloud solution, which is pretty new.
00:10:25And you just give it the docs, or what I like to do, I ask AI to do some in-depth research.
00:10:31I give it some initial docs and ask it to dive deeper, and then I ask it to build a skill which I can dump into the projects where I'm working with a certain library, and the results are really good.
00:10:42And for all these reasons, long story short, I think new framework and library versions still matter a lot.
00:10:49Or not just new versions, of course.
00:10:51Entire new libraries and frameworks matter a lot.
00:10:55But the reasons shifted, to some degree.
00:10:58As mentioned, things like performance and so on will always matter, because ultimately, you're hopefully building products for the end user.
00:11:06And the end user could be you, if it's an internal tool or something like that.
00:11:10But still, you want something that is fast, that doesn't crash all the time.
00:11:14So that all matters.
00:11:16But then, in addition to that, what matters today is how well does a certain framework or library work with AI.
00:11:25And that, of course, changes how these frameworks and libraries are developed or should be developed.
00:11:30And what the priorities of the developers behind these frameworks should be.
00:11:34They should really try to build something that forces AI to work in a certain way, and that minimizes the amount of mistakes AI can make.
00:11:43Because, ultimately, what we're, of course, kind of trying to do here is we're trying to force a non-deterministic thing, AI, into kind of a deterministic world, to get the best of both worlds, to get the flexibility and speed of AI, but avoid it going totally rogue and spitting out horrible code.
00:12:03And, yeah, new frameworks and libraries, I think, play an important role here.
00:12:09And, therefore, learning these frameworks and libraries certainly has changed.
00:12:14You, as a developer, will most likely not spend hours on a huge course to learn a new framework or library.
00:12:23You will probably, and you should probably, I certainly do, spend some time to pick up the basics with the docs, with a YouTube video,
00:12:33with a course, of course, that is also changing.
00:12:36I'm changing what I'm producing because I believe in order to efficiently work with something, no matter what it is, through AI, you need to have a basic understanding of the thing you're working with.
00:12:49You need to understand what to ask AI for, what to watch out for, potential pitfalls, best practices, and so on.
00:12:56So that is certainly changing, how you learn things and how you decide what to use.
00:13:00But that does not mean that the innovation around frameworks and libraries wouldn't matter anymore.
00:13:06I think it matters more than ever because I think there is a lot of empty space and room for innovation, for frameworks that give AI clearer rules.
00:13:16And, of course, always for frameworks, libraries, tools that allow us to build faster, more performant, more error-free, and so on, applications.
00:13:26Because ultimately, as I mentioned, that is what we all should be trying to do with or without AI.