Are Frameworks like React Redundant in an AI World? - Paul Kinlan

Paul Kinlan dives into web development's evolution, balancing frameworks vs. vanilla coding, the role of HTML/CSS, AI's impact, and the rise of disposable software. He also shares insights on team training and efficient workflows.

----------------------------------------
Welcome to WeAreDevelopers, the #1 developer community in Europe!
This is your one-stop destination for the latest tech insights, tutorials, and career advice to elevate your developer career.

Stay updated Dev Digest, with our weekly newsletter featuring the most recent tech trends, career guidance, and original content crafted by developers, for developers. Subscribe now

Interested in advancing your career? Browse through our job board featuring over 190,000 jobs. Unlock job opportunities with a free developer profile

Don't miss out on the annual highlight of every developer's calendar - the WeAreDevelopers World Congress. Network with 15,000 peers and learn from over 500 speakers. Secure your spot and save 10% with "wearedevs_yt"

#CareerInTech #Tech #ProgrammingTutorials #DevRel #CodingTools #TechJobs #TechTalks #DeveloperSkills...

Transcript

Welcome to another coffee with developers Today. I'm very happy to have Paul K. Kindlin with me. Paul Kan is a lead for Chrome Developer Relations. We've been crossing paths for decades because we, I worked on Chromium stuff as well with Edge, we fug up with Firefox and yeah's.
It's been both big proponents of the web and of vanilla stuff that is on the web because we worked with the tools that display those things. So it always been bizarre when people say like, oh, I'm a react developer and I'm like, no, you're actually an HTML and CSS developer because that's what comes out at the end no matter what you start with. So that's a bit that I always find fascinating that people think like they don't build things in vanilla, but they do, but with like 15 steps in between. Why I wanted to talk to you today is that you stirred a bit the pot with a pot with a blog post that you said the other day that, where you basically said like, you can see that in the nearer future all these abstractions, all the frameworks are not going to be necessary anymore, which is of course what would people fight over then? That's just unfair.
So that's one thing to worry about here. But why do you think that is? Why do you think we actually will go away from abstractions in the near future or the products will go away from them? Yeah, actually this started when, you know, I've just been building a lot of products recently and you know, one of the things I found like is, you know, I've been a web developer I know there 30 years nearly or 25 years. And you know, I'd always hop on to kind of the latest tools.
I'd still prefer vanilla as the kind of the way I'm developing kind of code, my output and everything that kind of comes through. But you know, being a manager, because I'm a manager of a pretty big team in Chrome on our side and I just don't get to build that much at the moment right over the past couple of years. And you know, the majority of my job when we do interviews with people say, well, what's your day like? I'm like, well, email is mostly my day and calendar and meetings and stuff. But you know, I started looking at chat GPT when it first came out, kind of amazed by kind of just some of the stuff that could build relatively simple things.
You could output a web application that would work like to do lists. Okay, I get it. Because to do Lists of everyone's favorite web kind of starter application. And then you progress and there's more and more tooling come through. And I've been using this one called repit.
And I was just kind of amazed by how productive I could be as a person, as a developer. Even though I'm still a manager of this team, I can build the things that I want to try and build. And it doesn't matter necessarily if they were just for me or for the team. And so I got into this bit where I'm just like building more and more and more things. And it just kind of hit me that I was just like, oh, do these machines?
Do these tools need like the kind of the abstraction that humans need? That was kind of the point I was trying to get to with the article because, you know, I've met with so many developers. Like, one example I've got, we were in India and we were speaking to kind of a company that was doing a video site, and they had to hire like 300 engineers, like in the space of like two months. And you. We were trying to convince them at one point to use web components.
And they were like, hey, cool, can you train all my people to use web components? And we were like, maybe not, right? And so like, kind of understood why they were using abstractions because they're there just to help the teams build faster, you know, more consistently, fewer bugs. Right? They've got it.
They've got these deadlines that they need to meet and kind of certain quality bars that they might have to hit. And then when I was building these with these tools, I was just like, it didn't seem to care. Like, you know, I could just put output like, hey, I need to do, you know, form validation, or I need to do, you know, spam bot detection, like all these different types of things. It's like, okay, cool, yeah, I'll do that. And it just outputs the code's okay, great.
And maybe it's uses some libraries, which is a little bit different type of abstraction. But, you know, it was kind of. It was doing all the same things that I would find that I would normally use something like REACT or REACT for, you know, you consistent DOM updates. I go through it and I'm looking at it, I'm like, hey, I'm actually the person who's. I'm using REACT to help me kind of understand what's happening.
The machine didn't particularly care, and it got this consistent output that I needed. And it was just me explaining kind of what I needed. And the goals I needed to achieve. And so yeah, that was kind of where the article kind of came. I'm wandering around now, but like that's kind of where the article kind of came from it originally was because, you know, I just don't know where the future of software development is going to go.
And I can totally see a world where these abstractions that we've built to help teams and individual developers create software might not exist right in the future. And so that's what was kind of pushing out with that whole process because yeah, it was just nice to see that everything was just kind of plain and just kind of ordinary and readable and I didn't have to learn anything else that kind of, you know, it didn't say that, you know, I'm going to use a React brand new version I don't quite know and understand just yet. It just output plain basic code. I was very happy when it did that. I mean, I mean that's not surprising because I mean in the end what runs in the browser is HTML CSS.
Even JavaScript generates HTML CSS in the end. So that's basically when you start with that you already skipped to the end, so to say. And it's not surprising that it also learned from these things because they're much less fluctuating. Like when, when we started, the CodecX was the first model and then lots of, lots of other code models that actually, these systems use. They actually use source code that is readable to them, which is basically the thing that comes out at the end of the abstraction, not the abstraction itself.
So I'm not surprised about this. What surprises me more is that we, it seems to put a kind of a light onto the whole concept of saying you'got to be, you're more effective and you have fewer bucsks. That's a good point that you said like with abstractions. But the dream that we've been chasing, which is like developer efficiency by going into abstraction by extraction. Extraction seems to be going full circle.
We basically by having to learn like five different abstractions. As a beginner trying to find a job right now, like company A wants React company P wants White company See what's htmx and you're like just learn the basics and then the other stuff will come to a degree is something that is, that is interesting to see that we, we focus on the wrong bit rather than focusing on the outcome of it. We focus on how it's being built and how much we, how much time is being spent on It. Right. I think that's just a nature of people.
Right. We do like to kind of, you know, focus on the tools that people use and like how people do it. Like, you know, there was a Chrome dev summit that we did years and years ago and we had all the BPs that these really senior people on stage and it was a Q and A. And like the first question was just like, which, which, which editor do you like to use? But that was it.
Like you had, you had the opportunity to ask kind of these senior leaders, like, what is the state of the web? And it's just, you want to know kind of how people build things. Right. And I'm always into that, of that type of thing as well. Right.
But I do wonder whether you mentioned about kind of like knowing the output and the output is the focus. Right. I've been noodling around with this in my head where it's. I don't know whether we're a kind of a spreadsheet moment, right. Where you've got like this new piece of software and these new tools.
They were primarily meant for, know, accounts, you know, small businesses kind of doing basic kind of tabulation and everything. And then you, you hop forward 30 years and all of a sudden you've got entire industries with processes and applications built off the back of this relatively kind of simple way of constructing code and computation. And you know, one part of me is just like, well, it's different with these large language models and these tools that I've been using. It's definitely different. But at the same time, you know, are we going to enable a whole bunch more people to build software?
And I've got this. And we. Off screen, we were talking about kind of they throw away software, I suppose. But you know, like, there's this bit in my head of like disposable software where it's like I can write something just for me. And because it's so quick to get something produced, I don't care whether I'm going to get a thousand users, ten thousand uses, one hundred thousand users for it.
It's just like something I've needed to get done and I can do it and I can just like type in kind of what I want to see and get something and then. And you twiddle about with it, get to kind of see how I can, can change it and kind of, you know, alter it. I don't know whether we're at that moment and you know, as soon as you start getting more and more people building those types of experiences there's only going to be a certain set of people who actually really care how it was built. Right. And so like all the tools in between.
I don't know where we're going to land on this thing for me. And the one of the reasons why I like rep it and m definitely not a spokesman for rep it but it's the favorite one I've used is it has this kind of notion of progress that you can see the diff as it comes through. I definitely wouldn't put any of my friends or my family into this application start off with because you have to understand the code it's outputting. And that's the thing for me which has worked well because I can see where it's made a mistake. I don't know what's go goingna be like in the future.
Right. O but that's the whole rag approach anything, isn't it? That's the whole rag approach like repeat and ask and do refine things. My worry about this and I Talked about this 2018, I think in a TED Talk where I said like back then I said that apps are the biggest step back in software distribution because the web you basically went to a URL, you used the thing, you discarded it, but you actually changed it to your needs. You needed a bigger font and it would display a bigger font because your browser did that for you.
An app controlled everything for the user. I could not take away any web content from people because it gets cached. I mean Google stopped caching now, but the Internet Archive still does. So an app I could say like I don't want this to be released anymore and I can take it away. A flappy bird could be taken away from people because it was an app as a website it would have lived on because there would have been so many copies.
There's no chance to do that. So that's what I what I was criticizing back then. But even more so what I'm confused about is that it actually that everything has to be an app for people to think that they're doing it. I had another blog post which I'm goingna put in the notes here as well where somebody else is using Brept and he was basically super excited about as well because it was so easy to build that app for himself. And then he downloaded the app and then he used it and I'm like ``mm that whole generation of that app is absolutely unnecessary in this case.
Like why is that a binary? Why that is being. Why is that being built as A code thing that you put on your machine then when it could just be the answer. So that's of course what everybody outse streams of liked. Everybody.
I mean you're at Google, I was at Microsoft. They all dreamt about that meta app that basically everybody uses, much like what we chat in China. That is one app to give you all the results that you do. So if our end users only need results that focus on AI building apps for us that basically then get not used seems to be wasteful to me but it seems to be what we're all focusing on right now. It's like everybody talks about agents that build apps for you and you're like what's the app for?
That's the question then is the app the real use for that? Or could that be just, I don't know, like a hosted. Hosted method or something like that on the cloud? Yeaheah. Yeah.
I don't know. Like I think that's the interesting bit that we're going toa find out right Like I am, you know, so one example, one of the apps that I built was just a time until like how like a countdown app. And so you know, I'm taking my parents to Japan and you know've saved up for years to try and do this and they've always wanted to go and so kind of doing it this April, it's going to be great. I hope it's going toa be great. It's go going toa be finen toa be great.
Great weather in April. Yeah, yeah, it's going toa be nice actually. But you know, I didn't want to use any other app like and my mom, my mum's one who's asking for this. Everyone uses that example, you know. You know, is your mom going to do it?
It actually, it literally is for my mom but she's always asking me like how long is it and how many days, how many weeks is it until. And then you know, you know, I can build an app that does that. And the reason why she kind of liked the app versus say you know, just type in a prompt every so often is just, just clicks on the button and it just does the thing. I do think people like that type of stuff and like the idea of, you know, when we were thinking about progressive web apps, you get something on the home screen and get it launchable. Like that's what I think people like I don't necessarily know whether people want the installable bit or like they like the long term kind of like hey this is always, ever present on my device.
But I, I do also wonder how many people actually want the. When it's more local, it seems more private. Right. I do think there's a kind of a like concept around that, that kind of thing. Whereas sy somethings on the web, it's like well, someone's tracking something or someone's doing something, right?
And it's like there's a bit of worry around that, but it's a bit of security by obscurity, isn't it? Like you basically say like oh I hated that it's on the web and everybody could do something. But that random binary that was generated by me probably has no malware in it. Right? So we don't know that either.
That was also the point of the other person when he wrote in his blog post, it's like, oh, of course there's a website that can tell me like I don't know what he built like some Bill Splitter with three people or something. And like, like yeah, it's called a calculator. It's on every mobile phone and it's in every operating system. But it's also built into Google. I mean I really enjoyed when both Bing and Google and all the others also became like this.
When the search bar was not only a search bar but also a functionality thing. Like when, when you, when you put in what is 3.5 and it would answer you 8 rather than giving you a website that is like a calculator where you have to type the 3.5 in again. But it seems like this has kind of gone away in the last few years and I guess it's because ads have taken over as, as search results rather than like giving you these kind of immediate things. The calculator bits in Google or being I have myself have used myself much either anymore, but they're still there. But I think there was a findability issue.
People just didn't think that that would work. Yeah, I don't know. I don't know what the. I don't know what it is. Could be a load of things could be regulation.
Who knows. That's why, that's why I love the machine learning in in device and also the hidden machine learning stuff. Like a Spotlight in Mac has so many cool things that I didn't know about. You know you can type in show me files that are five weeks old and our PDFs with like a background color of green and it would actually do NLP on that thing and show me those files. But I don't expect this as an old man of computers that had to tell them everything should be done.
It feels like this magic is hidden somewhere and people don't realize and they think they have to repeatedly built these things. Which of course brings me to the cost of it like the cost of generating those apps in terms of energy used and human time used as well. Like, I mean I love it that it brings people into development but I also wonder is like is talking to chat GPT for 25 minutes so you learn how to do an algorithm really as good as just looking into stack overflow and read the first three comments explaining why that thing is good. Every company said they're going to be green by 2027 or something and none of these say these messages anymore because all this j aai stuff is super expensive. Yeah, I mean just to be quite blunt, that wasn't on my mind when I was using this tool, you know, and I was just.
I was seeing my own productivity, right Skyrocket and there'things I could do u you know, it. It's not mentioned anywhere. I know that, you know, Jeremy Keith kind of brought it up from a conference that he was at recently. Like no one mentioned it or maybe one person mentioned it kind of in passing. You know, I think we probably do need to do more kind of public awareness around these things about kind of, you know, energy consumption and maybe add don on t the data usage on a bunch of other things on that side.
I think I just'that early on that no one's really been looking at that piece. I think. And I see this as a. This happens a lot in technology, it seems where it's like you don't understand the actual impact and the risks right early on you see the benefit, but then, then you have to work it out later, I suppose. And so, you know, I'm kind of glad that people are pushing on it.
I just didn't have any concept of it when it was kind of coming through. And I still don't write like, you know, you go to rep its website doesn't tell you. It doesn't tell. You know, you have to kind of infer what model it uses. I think everyone's using Sonnet like from Claude or Anthropics Sonnet, but you don't know.
Right. And like I can't remember the name name of the person I met a while ago. We were talking about kind of web sustainability and energy usage. Like even measuring the usage of like a normal website is heavily fraught and you don't understand exactly where like it's hosted and the energy consumption and all that type of stuff. And so the best unfortunate that we're doing is that we're making guesses right for even just traditional software.
And it's, it could be a lot worse for this you this type of technology just because of the power demands of GPUs and TPU'and I guess with code generation we're even on the, not even on the tip of the iceberg. I mean like image generation, deep fake video analysis, these kind of things. I once u in a newsletter I keep collecting these things and I show them out there and I have like generating one image is basically the same as charging your phone in terms of energy consumption. And like Maicrosoft Copilot always gives you four pictures all of them broken in a different way because you always have to generate like eight of them before, before the first one is correct. So that's a really, really heavy toll.
But at the same time it's of course exciting. It's great. It's like Star Trek talk to your computer and you give it back to you and u the, the it's interestingus A lot of times the apps that you talked about is what we ask people in job interviews how they would actually build an app like that. And now we actually asking the machine to do the same thing. They're going to hire them next.
Yeah, we'll see. So yeah, I don't know. It's a hard one. But I do understand the challenges that people have and we just got to stay on top of it. I think right now I don't have a better answer.
Unfortunately it's a definition thing that has to shift now. I remember when like Alex Russell and Francis Berryman were defining PWA and basically saying like what does an app really do? What makes an app an app and not a website? And I always broke it down to like it's a thing to do us to do a thing with. You know, it's an interface to us to a scientific problem or to a calculation problem or to a generative problem.
And the apps that we generate with these repid things don't have an interface. They most of the time are just a form. So the real impact I think of apps when it comes to making people more effective is having the right interface and best, I think the best example was Google Maps. Like the first time we had moving maps that you consume in ye I remember that actuallyah not like MapQuest where you had like arrows left and right and up and down. And all of a sudden people could go on, on an adventure.
They could go and actually inspect the data and find information in it the way they couldn't. And that's where UX and UI comes in. And that's why I think not all apps could be generated in the future, but it will be that these throwaway apps that nobody needs, the click through websites, you know, the things that we always be promised with what you sees, what you get, editors, we could now generate with those app generators. So that would be an interesting concept there. Yeah.
Maybe we'll have tools just to read them automatically and then red display them a different format. I don't know. Like, I think that's my. One of the questions that I've got is just like, where does this go? And I could see a world where, you know, what we know is the web is apps general.
It doesn't even have to be just the web, but the web'kind of my medium that I work on, right? You just, it does fundamentally change. And if you're defining a goal, need to get something done. You know, I had this thing, it was Ben Thompson and he was talking about the Vision Pro. And there's no way that you can create these kind of immersive worlds, automatic, like manually.
You're gonna have to have some sort of generative kind of technology to create an immersive world. And like, he's kind of then moved on to this idea of, well, UIs will just be dynamic, right? Like, you know, they'll know the goal or you'll express the goal that you want to try and achieve and then the software will just build the ui, but to achieve that goal. And it might only ever do it once, right? Because you only ever need to do that task once.
And like, geez, I don't know. That does just raises the question about like, what is the, what is the future of any application or web app or website, right, in terms of, you know, how we using them today? Like, a lot of people build applications, right? And they build it to do a function that kind of optimizes some process that they've got or make something a little bit easier. And it's like, oh, okay, yeah, these tools can start to generate them.
Like, what does that mean, like in the future? I don't know. Right. And I'm just like, I'm sitting there going, you know, I'm lucky that I'BEEN around here for 25 years and I've, you know, got to this point. I don't know what's going to happen next.
I give a lot of talks about June junior developers right now and what they need to what they face and I'm very happy I'm not a junior developer right now because you have all the opportunity in the world, but you have no opportunity. It seems to make your mark like you ve seen as like somebody who uses something to produce something. You're not seen as somebody who's like the creator anymore or the. The craft of software seems to be going away much like the craft of filmmaking with like AI generated CGI stuff in movies and like writing and like blogging and like''it seems like we're with the AI slop, as Simon Willilson puts it, like going everywhere that like AI is overwhelming everything by generating lots and lots of stuff. I think there will be a backlash that people want to get like more handmade software kind of.
Yeah, yeah, yeah, yeah, yeah. You know, more things, like handmade things. But I think there's definitely a space for kind of craft and curation and you know, human interpretation on, you know, you know, expression through kind of tools and stuff. I don't think that's going to go away. I don't know.
I just, I just don't know. And I do worry about junior developers and people coming in right now like might. My eldest son, you know, he's doing a history degree, he's just gone into kind of university and you know, we're trying to talk to about hey, what do you think about you might be doing? Doesn't know which I think is a lot of kids right at that point. But I also don't know the world that he's going to come out into.
Right. And so yeah, it's just kind of, it's just kind of weird right now. And I think there's a bit here with, I think maybe you're on the, I don't want say maximalist view but like AI definitely is going to change things or it's not right. And I'm on more of the side that there is a big change happening but I also don't know how to deal with that right now. Well, I like, I liked GitHub's announcement when they're like we're going toa make a billion more developers and that basically ye.
I think that's, that's where, that's where old people and people who've been around for a long and stalwart people have to start thinking about we need to redefine the job titles that we have as well. What is a developer. What is, what is a software creator? What is that? What is an engineer?
Like? I mean there's always been a sliding scale. Like I did the same job for 25 years and I had like 15 different job titles. But there is, there are certain expectations with a certain word and developers seems to be quite free of values in terms of like o like free of. Anybody can be a developer.
Nobody's going to be worried about taking it away. But if I said anybody can be an engineer, people would be like I went to university for eight years for that Note. Don't tell me I, that everybody can be an engineer. Yeah, yeahah. Well I use ChatPTT.
I'm a doctor now. Right. I can diagnose any problem. So I don't know what's go goingna happen. Well, I mean that's like, that's like webd.
Whatever you had, it's going toa turn out cancer. Like you know, like that's, that's the question again. Like also the quality of these things that get generated if these little apps that you talked about that, that is basic knowledge that has been solved so many times. And again, why did people start from scratch every single time? Yeah, that also brings you back to frameworks and abstractions.
I mean we always talk about, don't repeat yourself but then you, you see the, the rise and fall of these, of these frameworks is always the same story, you know, Yeeaheah, it's more effective, it's shorter, it's smaller. Oh, now we need to add this, at this, at this. Oh, it's not shorter, it's not easy to start anymore. The other one is the black one that's better and easier. So I mean Ajax times when I worked for a Jackson or wrote for a Jackson, we had like 167 different AJAX libraries and in the end jQuery was the only one left.
But yeah, it'it made us more effective. But I, I really like the idea how you actually realized like abstractions are for people and frameworks are for people, not for computers. So computers generating code doesn't have to be readable, but it has to be effective. And that's where I think it will be interesting to see what's the future for frameworks in thise. I do too.
And I feel like there's a bit here'like there could be just frameworks that are more kind of friendly to these types of tools. Right. And so then there's just an abstraction that the actual tool uses as well. We don't know at the moment. But, you know, there is a bit where it's just like, I do think there's some fundamental pieces that we need.
You know, one of the things that I did find when I was using these tools was, you know, the date cut off. I suppose to some extent, if the technology isn't as pervasive in the ecosystem, then the tools won't know how to use it. And so I think one of the examples was things like view transitions. You know, there's a definite kind of, you know, I actually like the kind of the aesthetic, UI and UX kind of feel of applications when they feel trans. Like they just feel quite cohesive on the navigations, feel cool, feel good.
Like the tools today don't really know how or seemingly right now don't know that this as a technology exists on the web platform. And so there's a bit here of just like, I can totally see there's a world where, you know, human, humans are a little bit more kind of ahead of the curve, can keep on top of stuff. The tools right there, right now aren't as capable in that sense. And then I do wonder, yeah, about kind of what will be this next generation of tools that come out and how will they be trained? Like, that's just a big, kind of a big question.
It is just like, you know, HT Max comes out and like there's only one or two apps that use it to start off with. Like, how does the tool know right now how to use. I don't know at the moment. And so, and then DevTools couldn't do them because it doesn't generate the normal DOM stuff and there have to be extensions. But dev tools again for different everything Sing one of those.
I remember when like prompt engineering became a thing and people like, okay, so next will be done optimizing. And then I saw frameworks for a prompt engineering that actually takes human sentences and turns it into an optimized prompt for chat GPT. So yeah, we always find these libraries to fix these kind of things. So. So as you say that.
So actually one of the things that I struggled with repit was it didn't, didn't do my form design very well. So it didn't take advantage of, I think maybe accessible forms and didn't do password reset properly. I had to kind of teach you how to kind of do that. And then, so I was kind of, as you were saying that, I was just like, yeah, yeah, there's bits here where it's just like, you know, I was thinking about like, well do I need another tool which will help me take my two line prompt? Because that's all I could put in to start off with.
Second you press Enter, it goes off and build the app. I don't want that. I just want to be able to kind of type out my spec and then I want someone to kind of come in and like say well actually this spec doesn't define the security requirements or the accessibility requirements or the performance requirements like the things that you would expect from a traditional application specification. And so I'm just like, you know, I am actually going to build that tool that you just mentioned which is just like yeah, we need to kind of take the prompt and then do more with the prompt to make sure it's go goingna meet all the legislative requirements. I as a web developer know that I need right now because I don't necessarily know whether they're in the actual kind of tools unless you prompt.
And so yeah, there's a bit there just like all of a sudden for me like the person who has the idea, the product manager, the design, like all of a sudden they're way more in control I suppose of kind of the destiny of the application. And yeah, so as you said that I was just like yeah, there's a bit I need to build a tool which is actually going to refine that prompt because. Because there's potentially loads of things that are missing like the application. When I first built the one that one of the really big ones that I built it didn't do kind of, you know, spam or denial of service detection and let you know, as I built it I was like, you know, if I spent more than the 10 minutes thinking about it that I did just kind of write that two line prompt, I probably would have thought about the user registration and what the implications of what happens when the user registration can get spammed. Right?
Because that's what a business does, right? When they're actually thinking about their applications. And I'm just okay. Y. I think Gith is working on GitHub is working on some auditing tools that actually does these kinds.
Really gener for the generated code, not for the prompt. Director's great with the prompt. We probably would overload the machine. I think when you also talked about few transitions. The other problem of course is that retraining and re weighing models is really really expensive.
So that's why all of this stuff is rather outdated. So basically there's two year old code, five year old code at times in there. I Mean, I've seen ANSW is from Stack overflow that are 10 years old that basically still say this is not working. Which brings me to another thing we can definitely talk for another hour about. Not today, but I think we did a.
We did with Viewue transition, a great example. I love them to bits. I think this is exactly what I want to have. I want to build a website that looks like an app, that feels like an app and I don't have to do it. Like we tried it with amp, didn't work and.
But we view transitions which kind of go back to the Internet Explorer 4 or 5 transitions that it had back. I remember that stole from PowerPoint that basically are there again. But whenever we release these things and both in Microsoft and in Google and everybody else was Mozilla even more, we always are playing too nicely. We're always saying like, oh, this is a preview and you have to go to your browser and set this, go into this origin trial or set this flag before you can actually use it. And it only works in nightly right now.
Whereas native code says like okay, it's gonna be an X operating system and you don't ever a say in it. So that's why a lot of people get like stop trusting new technology web platform features because we said immediately that they're just in beta right now. We want feedback from you, which we never get in the droves that we want to have. Like it's when I, when I saw like 90% where other engineers from other browsers giving us feedback and not people outside because they don't have time to start playing with it. And then of course these machines will never learn those things.
I think the getting new technology for the web platform out before it actually people start using it and trusting it. There's a half year year delay in that one until people go for it. And I don't know what it is until. Until iOS Safari supports it. It's not available to people I don't know.
But it is a problem that a lot of technology was too timidly rolled out I think. So people didn't trust me. Yeah, I mean on the web though, it's like's I get it though. But like there's a bit here of there is no one owner. I know Chrome is probably one of the most used or Chromium is one of the most used kind of browser platforms right now.
Like the power of the web is that there isn't one owner that can make those unilateral decisions and the trade off is that we have to accept that, you know, development might take longer or kind of pace of change of output. When people create things and get it out to everyone, it might take longer. I think that the one benefit that iess what's the big benefit is like it's the link still, right? And like all these are the platforms for me this say I'm going to rightnt now. But like all the other platforms, they all want to ape the web in terms of hey, like quality and maybe they do, maybe they get better quality.
There are very few platforms that I'actually got the same ability to reach everyone with just a link. Like you know, you look at kind of the Play Store. I think Apple's done it in the past as well. They want to have like instant installable apps. Like I just don't.
I personally just don't ever see these ever taken off. Like the power we've got is the link. You click on it and it goes to a thing and you start to run it and use it or read it or play with it, whatever you kind of do on the, on the web. And like there is a bit for me there of just like I definitely want us to keep. For me, I want us to keep that I the bit I was going to say on top of these things within, within newer technologies and if more and more people use these tools, there could be an argument that we regress in terms of quality.
I don't think that's going to be the case just because if you look at something like baseline as a kind of a concept where it's like trying to just make sure that people know what the web is possible of. Today the web is incredibly capable across the majority of major browsers. And like we just. What we're struggling with right now is that people don't start to use those technologies because they don't know that you can use those technologies or you can rely on them across your customer base. And I think there's a bit here just like it's easy to point at the Safari team and say, well, you're slow, right?
Like you'you know, you're kind of holding things back. I don't think that's case. Yeah, is fast. So yeah, WebKit is. It's always there.
They're always happy to bring things in and maybe to cap this off, people are going to use these tools. Maybe the benefit of the actual people with the experience is like if the tool doesn't have an implicit knowledge of say knowledge as it's a human concept but like it doesn't actually understand kind of hey view transitions. You as a expert can say this is the view transitions kind of definition of the API shape. Use this and it for a lot of different types of APIs that actual prompting. And given that kind of one shot, two shot, whatever it is kind of extra context really helps.
And one example that I've got is I didn't, I was using Flux I think which is the image generation API. It only has an open API spec, it doesn't have a Python library, a JavaScript library or anything that I could find. And the tool rept definitely doesn't know about it. And you can just say, well actually just drop. Just use this open API spec and then it works out how to actually use that open API spec.
And it's a mult multi step API. It's not like one call does one thing, you have to do one call and then poll and poll and poll and it got it straight away like the tools actually understood. The more that you kind of drop in, you'd be surprised by how much you can actually start to kind of take advantage of the higher level like these newer technologies, whatever. But you as a person have to be the person providing that information, right? So I just thought that was great.
I think the problem though is that next time you start with it it will have forgotten about it because the original model under it doesn't have it as part of it. So that feedback loop doesn't work right now. And that's really frustrating because you have to start from scratch every single time with that one. I think there's also the problem that there's far too many don't use this caus, it's not ready yet blog posts out there and a lot of AI bots actually get trained on these things as well and that's why they get don't get up voted on Stack Overflow and that's why the Stack Overflow winner was five years old and not like one of the cool new things that might have one little feature that is not working. So that's, that's a big problem as well that we, that we're telling the machines that.
I mean there is, there seems to be a whole movement not anymore as much in the last few years there was just so many ranty block posts about saying like oh I used this thing from the web platform and it's still not fixed after so in so many years. And you're like well where any bugs filed or anything and like how do you expect it to happen magically. Like we both know that it's hard to get time for engineering engineers on browsers to fix things that are in the platform if they're not immediately. And if something isn't used, we don't get it on top of our buck list. So more usage, more opportunity to fix it.
It's a very simple equation. But we didn't see that. And that's why those machines will never learn about these things if they're not being used, right? Yeah, but it's fantastic. I mean that's why it was fun to work in devools because you're basically you're in your own environment and you can use all the cool stuff that later on you can roll out to the web like that was.
I have a question for you. Like how do you see devools changing in this kind of. I know that you. It's more in the past but like you see devtools types of tools changing across the board or like new typess coming out. Definitely.
I mean I'm another company that wanted to hire me but couldn't afford me because they're in Singapore. The German work contract was out of bounds for them. I'm really excited about they doing a Figma plugin that generates react components from Figma by doing natural language processing and also image analysis on your stuff that they've been doing. And they did that for like 15 years and before machines were fast enough to do it properly. So instead of just creating a thing in figma and saying this could be a button or could not it actually generates the button because it realize it looks like 5,000 other buttons we found on the web.
Right. So I think this immediate tooling from one place to another I think is go going toa be more interesting. There's of course linting and checking and these kind of things are are not changing but a more an immediate like what you see is what you get that we always dreamt about is I think much more in the nearest space that we have right now. And of course the testing things, I mean I'm speaking on Thursday I'm speaking at. At an AI summit for.
For qa. People like accessibility testing with AI and AI generating accessible interfaces which is not what AI should be doing. So these kind of things like automation isn toa be a big thing right now. And that's where I think the. It makes me wonder if we in the chromium devtools have gone far down, far too much down the rabbit holeer.
We think people know what we're doing. But when you looked at the stats, everybody used the console, the elements tool, and. Oh, no. Yeah. No, no, no.
Yeaheahe similar thing, just like you. There's like three panels that everyone uses, one panel in particular that everyone uses, and everything else is a little bit further down. And. Yeah, I think a lot of people just want to be fixed. Yeah, just debugger, you know, like, debugger gives you so much information.
What do people do instead? 5,000 console log statements, because it's easier. See my code print. F. That's how we learn it in school, right?
Like, yeah. Well, this was great. I think we could do this more, and I think we should, because this is working really well. But we don't want to. I mean, I.
I always wonder about this. I mean, I'm not a podcast person. I don't drive well. I drive car a lot, but with my partner, and she hates tech stuff. So we watch, listen to the radio.
I mean, I see podcasts of two and a half hours, and I'm like, who's got time for that? Who's listening? Stu O. I do two walks a day with the dog, and so they're five minutes each, and so I get through quite a lot there. Um, it goes different blends, though.
Like, I do like a lot of Ben Thompson's one as well. Like, he does 15 minute ones, which are great for certain things, like on the toilet or something on or whatever. And then. And the longer ones are like, I can just like, do across a day when I'm walking and stuff. And so, yeah, I like long ones and short ones are great.
I think it's just a great. I just think it's just a brilliant medium. So if you want us to talk more, if you like this, please tell us about it. And I'm not going to say, like, and subscribe because we don't have those features and we don't make any money with it either. But I think this.
This is fun and I want to do more of those because it's just great to chat about. And I mean, we're both super honest about it. We don't know and. But we both. We both interested in.
Because we've seen so many things moving over the years and. And where it's going next, we don't know, but it would be interesting to try different things out and see where it's going. So that was great. Thank you so much, Paul.
All right, thank you. See you soon. Bye.