Frontend after AI: The New UX

Future Frontend 2026 / 32:55

Watch this talk on YouTube

Transcript

36 paragraphs

This is an automatic transcript of the recording above. It is published in full and unedited, apart from correcting names the recogniser reliably mishears. It will contain mistakes.

00:05Hey, thanks, man. He said that book was like a best-selling book. Fun fact, the book actually came out of that video uh that your friend watched. So, yeah, that's how it Anyway, uh I'm done This mic is moving on its own. I'm uh done flexing. He did a good job. Uh as he mentioned, my name is Tejas. That's pronounced like contagious. Um and I have like I have This is a like 50-minute talk, but I have 30 minutes. So, we're going to like fire [snorts] through it. It's going to be fun. Okay? Uh in fact, I think it's the last talk of the day. So, we're just going to like have fun, you know? I think you you've learned plenty. The past few spea- the speakers did an incredible job, uh especially like um the ones before me. And so, now I just get to come here and and and we can vibe a little bit, you know? It's the end of the day. Let's have some fun. Um speaking of fun, um I >> [laughter] >> I work at How many [snorts] of you associate IBM with fun? Anyway, um I I work at IBM, uh 115-year-old company. Isn't that something? Um and uh that's not what we're going to talk about today. What's more relevant for this talk is that I have studied human-computer action at Stanford. Uh and and boy, what a thing to study. I mean, nowadays human-computer interaction is is changing everything, right? Like the way we interact with computers, the interaction of humans and computers fundamentally different,

01:23bananas. In In fact, what I studied may be outdated. Or is it? And that's kind of the the point of this talk. Today, we're going to be talking about front end after AI, um the new UX. Because look this conference even, right? It used to be called React Finland. Anyone remember that? Anyone come to React Finland? Yeah, all all of you, almost. Right? Now, it's not called that anymore. You know why? Because front end is different. Like React's still the most used library for building front ends on the web, but who cares? Um in a way. Like it's used, it's a great tool, but front end is changing. We we live in a day where I don't write React anymore, but Claude does, right? Um anyone can build anything, more or less, or at least some sloppy approximation of what you want, right? Like you have things like Claude code and Cursor, Codex, everything starts with C, um that IBM we create our own uh fork of VS Code, it's called Bob, Bob the Builder.

02:21Uh no joke, that's actually a thing we make. Um similar to Google's anti-gravity, you know, everybody's kind of making something. So So now everybody can make everything, and it's What does that mean for front end? What is it? And that's kind of what we're going to talk about in this talk. Front end is still the front end, right? It's still the end that is the front. It's It's this deep stuff, man. It's still It's the end that people look at, the end user looks at. It's the tip of the iceberg before they get deep into your platforms. It's still the thing we present to people. This is why this talk is here, because it's it's an accessibility/UX talk that bridges the gap. Day two, tomorrow is very like MCP and agents and AI, and it's very hands-on. This talk is the bridge between day one and day two.

03:07Okay, and we're going to talk about about about front end through the lens of AI, um and also through the lens of UX, accessibility, and what I believe is the word of the the the season. It's taste. Taste is the thing that matters, arguably more than all other things, um nowadays, that anyone can build anything. Cuz the truth is, when anyone can build more or less anything, um it doesn't mean that all of it's going to be good. Uh you know what I mean? Does this make sense? Like for example, everyone people have thought, "Hey, look, everyone has a camera now. Oh, there's going to be so many photographers." Yet open Instagram. You know what I mean? Like it does Just because everybody has a camera doesn't mean everyone's an amazing photographer. We don't see high-quality, beautiful 4K HDR uh with taste. So taste is the thing.

03:52Everyone can generate code, but few can create beautiful, tasteful products. In this talk, we'll talk a little bit about taste, and then we'll make it practical with some hands-on stuff. First, let's define taste. Who Anyone feel confident to define taste up here? Uh me neither. Uh but uh I don't have to lean on myself. Thankfully, there's research that has happened for years, decades in the taste. And so, we're going to cite like peer-reviewed research from top-quality journals so that you don't get my opinion cuz I I Here's the news, everybody. I'm not a thought leader, okay? I can't lead anything. I just I just read. I'm a thought reader. Anyway, so um Wow, how do we define taste? We can define taste academically through the lens of science uh this way. Taste is a trainable skill. So, good news, you're not born with amazing taste. It's a thing we learn. Taste is a trainable skill where we're able to tacitly judge quality based on external standards.

04:47Wow, what a mouthful. Um Some Some of these words are bold, so we'll talk about what this means. And this comes to us uh from a paper published uh titled The Development of Aesthetic Experience and Judgment. This paper was published when I was born. And it And it um it does the best work of defining taste for back then, which we'll see together that translates to today. It's a trainable skill. This means nobody's really born with amazing taste. It just It's something that comes to us over time. And in this talk, we'll explore like how do we actually train it? Um Does anyone know what tacitly means? Tacitly means like you have a gut feeling. Like it feels right, intuitively. It just I just I can't tell you why, but I just know it. I just know that's a fake Louis Vuitton bag. I just know it, right? Um I just know this There's a statue um in the in the United States Museum of Modern Art, which is fake. And and the label says, "This is probably fake." They can't prove it's fake, but the taste is there, right?

05:41That's what tacit means. You You have a gut feeling. Uh tacitly judge quality based on what? Based on external standards. This is very important because as human beings we make tacit judgments all the time. Uh and sometimes those judgments of quality are based not on external standards, but on our own bias. Okay? Look at me. Yeah, you know how I look? People tacitly say I work in tech support. Uh that's that's a thing that has happened, right? And so tacit judgment based on internal bias is problematic. Um but tacit judgment based on external standards, aka science and fact, uh that's called taste. Uh when you can just tell, right? Let's Let's explore this through some examples in the wild.

06:25I think examples are incredible as sort of reference points. And so, um I think the best example of this is money. There are people in this world, not machines. They don't even trust AI with this, but there are human beings who it's their job to identify money. What's counterfeit and what's a real uh bill, cash bill, right? Um how do they do that? And you would think, well, they develop this taste for I know what real money looks like because they study all the counterfeits. Oh, the scammers, they use this specific paper that, you know, fake money has a smell to it, has a taste to it. They They You would think they spend their whole career uh studying fake money. It's actually the opposite. Uh they they actually spend almost all their time identifying and studying real money, valid money, actual issued from a treasury, so that when they see something fake they immediately know.

07:16They tacitly can judge based on what? Based on the external standard of real money. Does this make sense? Um that's literally how they develop that taste. This will hit closer to home. Um anyone ever work with a design I If you're not a designer, but a developer, you ever work with a designer who's like has got amazing taste? Yeah, many of us. My first job ever, uh I worked with a creative director. Her name was Stephanie, Stephanie Lupa. And she gave me this like design which required uh dangerous stuff, man. It required uh a div to be centered. In like In like 2012, you know, and it was impossible. And so, she said, "Okay, this is the design. I need you to do this." So, what did I do? I mean, we had flexbox, right? So, I put in a div, and I I still remember display flex, align items center.

08:01And And I was like, "Okay, yeah, looks LGTM, bro. Looks good to me." I I said, "Stephanie, come, it's done." She takes one look at it. 300 milliseconds. She takes a look, she's like, "That is not centered." And I'm like, "What do you mean it's not centered? It says align items center, Stephanie. Stephanie, do you even code?" And she was like She's like, "It's not centered." I said, "Stephanie, the computer doesn't lie." She's like, "I don't know what you're smoking. That is not centered." We went back and forth a few times until she was the boss. She's the creative director. So, she said, "It's not centered. Figure it out." And I said, "Well, Stephanie, CSS doesn't lie." CSS lied. Um It was So, I found out that the font uh apparently had like some weird line height uh that like [laughter] added space underneath the bottom of the text. And who knew, you know? So, I drew grid lines, and I made sure. Um but she had this taste, right? It's taste that was trained over time. Uh so, she could instantly know based on what? Based on external standards. She looked at hundreds of thousands of things, and she knew. Um One last example is um music. You know, you ever go to a concert Oh, there's metal concerts in Iceland, huh? Uh You ever go to In Finland, sorry. Damn.

09:10Kick me off the stage. Um You ever go to It's the end of the day, but whatever. Uh You ever go to a concert, and and somebody's, you know, they're singing or they're playing an instrument, and uh And I still can't believe I said Iceland. And And they Somebody's singing, and they hit a note that's off-pitch, that's off-key, you know? Happy birthday. Oh, Jesus, what happened, you know? And you you We all collectively do it. Like, "Ah, that's nasty." They missed How do we know? It's because we tacitly judge the quality of the singing based on an external standard being relative pitch, right? Like you you just know in the key of C with music, you can do G F A minor, but you do a G sharp major, you're going to have a bad time, right? External standard. And so you just know this. So, taste is defined as um a trainable skill. We can all train it based on rapid judgments even before our conscious brain kicks on, um but based on external standards, not internal bias. Difference between like racism and taste, okay? And so, why are we talking about this? This is supposed to be at the accessibility thing. Well, nobody's going to use your thing if it's distasteful. In fact, here's some research about this. Taste raises the perceived value of products and influences the intention of customers to purchase your thing.

10:27Apple's [snorts] event WWDC is tonight, and tonight they're going to show you some tasteful things, hopefully. They're going to zoom in all this this phone is made with glass and titanium. Ooh, and then they zoom in and they're like, "Look how beautiful." By the way, it costs like $10 million. Right? Right? And then we're like, "Oh maybe I You know what? I will pay for it. It's quality, right?" And so that's exactly what taste does. And they they know what they're doing, right? And it enables us to pay for things. Um so anyway, taste. Why are we talking so much about taste? It's the game changer for a world now of human computer interaction where anyone can build anything. I've said many things about taste. One thing I want you to remember strongly, write it down, get a tattoo on your forehead, is is that taste is trainable.

11:15It's the good news. That's really good news. It means we can all create wonderful things. But how do we train it? I'm here to give you the recipe, the science-backed recipe on how you can improve your taste if you if you feel like you need to need to. We all need to. How here's the recipe. Number one, immersion. You ever wonder who even makes taste? Like who who gets to decide what makes this like lady at Vogue decide what's fashionable and what's not fashionable, right? Um, turns out the people who have really amazing taste is just because they were born into situations where they had literally nothing else to do except look at a bunch of stuff and like make judgments. You're if you're born rich and you just look at dresses all day uh and your mom tells you that's a good dress, that's a bad dress, then you just know what good dresses look like, you know? Uh fun fact, so it's just like a privilege thing. Anyway, um so how do you get good at taste? Full immersion.

12:09You immerse yourself in articles. Uh you immerse yourself in in a in a system of valid cues. In the case of software, you immerse yourself in this beautiful application Linear, linear.app. Wow, wonderful. Or you immerse yourself in something kind of ugly and you're like, "Oh, that's kind of ugly." Whatever it is, you're deeply immersed in an environment full of cues. Great website, ugly website. Um, accessible website, inaccessible website, right? You're just constantly being met with these cues, but there's a final ingredient, which is you need corrective feedback and you need to be receptive to that corrective feedback to then refine your taste. Um, incredible work by scientists Dr. Kahneman and Klein published in 2009 in um in the American Psychologist, the Journal of American Psychology, uh titled Conditions for Intuitive Experience Failure to Disagree. And this is the recipe for fine-tuning our taste. We have to be deeply immersed, we have to have cues that we can respond to. We're not going to get it right all the time. We will receive feedback. Actually, this was not as good as you thought for this reason, that reason, and we use it to refine our taste. Couple of more recipes to help us refine our taste is um Dieter Rams, the head of design at Braun uh for a lot Anyone using Braun product Braun is incredible, they're known for their design, similar to like Dyson. Um, he

13:28says, "As you expose yourself to these cues, um, what you want to do is write down, like, this is what makes good things good, and write it down as an external standard, so then you can judge the things you make against the same standard." Really great advice. Um, in fact, this advice was so good, it was taken to heart by Jony Ive, the head of industrial design at Apple. Uh, I'm I'm holding something that Jony Ive started designing like when he was younger. Um, and and many of us, if you work on a Mac, you're using uh, Jony Ive's design via Dieter Rams, right? Um, everybody knows the story from Steve Jobs, where he talked about, you know, I dropped out of university, but I sat in this one typography class. You've probably seen this, and he talks about like, "Oh, I saw the font. Wow, what a beautiful I was totally not on LSD, but I saw this font, and and it was so amazing, and and that the design there cross-pollinated into Apple's sense of design, right?" And so, another good practice for us with, um, developing taste is cross- pollination, okay?

14:27Um, I'll say this another way. Taste doesn't just, um, judge better, it perceives differently. Uh, for example, you know, I could ask you if if if if I bring a junior engineer into the room, someone early career, right? And I'm like, "Tell me about the call to action button." They'll tell me about the call to action button. They'll say, "Oh, it's blue. It represents the brand well, etc." Um, if I call in somebody more experienced, they'll comment on the button, but they'll also comment on, um, this button actually on some devices renders under the fold. Damn. They'll also comment on on slow networks, this button is actually not visible at all. Like, they comment on the things around the button instead of just the button, right? When you have really good taste, you're not just judging the thing better in isolation, but you're thinking about the whole ecosystem. You're thinking differently, which is I think kind of a nod to what Apple was talking about all those years ago. Um, but anyway, taste.

15:20We've defined taste, and it's really the main difference in this era in the new front end. How did we define it? It's a trainable skill where we're able to tacitly judge quality based on external standard. Again, it's not our definition, it's from the research. This this is also front end, so web. What does this mean for front end? Well, we just this Guess where we are. We're in the block of accessibility and and and performance talks, right? We have standards in front end. It's so cool. We have actual standards. Input to next paint. How long does it take? Largest contentful paint. How long does the largest contentful portion of your website take to load? Cumulative layout shift. These are actual standards. With accessibility, we have Arya. We have the web content accessibility guidelines or WCAG. We have like objective measures of taste among other things. Here's another one that I personally optimize for. It's time to done. The truth is I don't really like using the web or front ends at all. I don't like using computers because the more time I spend on a computer, the less time I have to spend with my friends and family. I would much rather like go bowling with someone or or many people or watch a movie or something. Computers to me are fun to build on, but a distraction to do work on.

16:36I don't like spending time booking flights. I like going on flights, right? There's an incredible article by Google, old one. It was just published on web.dev titled "Milliseconds Make Millions" where they looked at how a 0.1 a 100 millisecond improvement of site speed increased conversion rates significantly. That's just publicly on the web. It's actually pretty old, but it it the point stands, right? And so on the web we have standards. We have numbers we can optimize against that already improve the tastefulness of our apps, but we may be able to do better with AI. And this talk is the bridge between today and tomorrow's kind of AI heavy section. So, uh the truth of the matter is I really don't enjoy using the web at all. Uh it's unpleasant. And what we're going to do now that we know what taste is and how to feel things tacitly, we're just going to spend some time um using the web and feeling some pain uh together if if your taste agrees with mine uh and then we look at some solutions to that, okay? So, we'll just do some exercises in taste together right now.

17:42So, um has let's just open literally any website. Uh let's open expedia.com. Uh I want to maybe buy some travel stuff or something. Um and am I on the The Wi-Fi here is kind of slow. Let me switch to my hotspot. This is already problem number one, by the way. Um this is actually very tasteful from Google, I must say. But anyway, um let's expedia.com. Immediately, I'm asked to sign in. I don't even know what you're offering me, but you're asking me to sign in, right? Let's go to I don't know adidas.com. Um The web is just so hostile. Look at like I nothing happened, but I'm being asked to do so I I really hate this Like And anywhere I go, um name one. I don't know. Um the coca-cola.com. Let's just Let's just Google Coca-Cola.

18:28I don't even know the domain, right? And even Google, Jesus Christ. I don't even speak German. This painful. Like I I do speak German, but like when you Like when you Man, I hate the web. When you When you open a browser and you go to some domain, you're leaving your house in a way and you're going to someone else's house where they've made decisions, they've made decorations, they've made decisions on your behalf. But the thing is they were not thinking about you. This This is not for you. This is not for you. This is for some legal person somewhere, right? And so, the web has become really hostile I hate this so much. Um in fact, there's an amazing website. It's called userinyourface.com.

19:09Has anyone seen this? Oh, it's absolutely fantastic. I think it's is a great representation of the tasteless web. Check it out. Hi and welcome to this low contrast nonsense. Um to to play the game fill in the form as accurately as possible. Okay. Oh, that's not what I'm supposed to click on. Ah, right. I see. Please click here to go to the next page. Okay, so I'll click here. There's a timer and a Oh my Jesus Christ. And an and a Okay. Choose password. Ah, it's not a placeholder. So I have to backspace. Um Uh hello is my password. Uh my email is Oh, this is also not a placeholder. So Tejas@ uh gmail.com I think. And then Oh, okay. Well, here we go. Um I don't know what's happening with this carousel, but whatever. Um I do not accept the term. No, I do. So I have to un Okay. Um cancel. No, next.

20:02Next. Next. I can't Why can't I I But this means I do not accept the But it says my password requires at least 10 characters. Hello, it's me. I am Tejas. So it has now 10 characters and do I click cancel? Are you sure you want to cancel? No. Oh, okay. Please do not forget to ACCEPT OUR TERMS AH! I CAN'T! THE WEB SUCKS! OKAY, LOOK. I ANYWAY, THIS IS MY POINT. This is like the amplified version. Yeah, but it's not that far off from like regular web use. Honestly, I I'm still The time is ticking. I don't know why. Um But anyway, I'll show you an example of good stuff on the web. Uh futurefrontend.com. Honestly, I love this website. It's so cool. Look, there's not a single cookie banner. Praise God. Um I'll go to the schedule and I just see things. It's so cool.

21:00What if everything was like this? How might we do that? Um Yeah, yeah. Exactly. Whoever did this >> [applause] >> I do this talk at other conferences and usually in this portion, I'm just like totally like destroying the conference's website cuz their websites usually suck also. But Future Frontend is genuinely good. I'm I'm quite happy. Anyway, so now I'll try and spin it differently. I left my house so to speak and I went to futurefrontend.com and I saw this and I was happy. It could have happened that I left this and went to some website that was hard to predict what would happen, right? What if there was a way to instead of me going to some other website, some other house, I bring the data to me.

21:42Ah, wouldn't that be something? I'm in an environment that I know and I trust and I love, built for me, and then all this cookie stuff, that's handled somewhere else. Well, this is kind of what agents allow for. And and let's let's explore this with taste a little bit. I could for example Man, I need to be quick. I'm running out of time. I could for example do like I built a little local agent here in my machine, nothing too fancy, and I could be like, "When is the speaker dinner at Future Frontend? Check my Gmail and get and and how long would it take from the hotel, right?" I could do something like this. And I just I wrote this myself just like in with with Node.js. Like it's not really it's And so I'm hands-off.

22:30Open my email and just do stuff. Please read all my emails while you're at it. Um Anyway, and and control it and it's you know, it's going to do its thing or it's going to like search here. We just follow the log here. Take a snapshot, search for Future Frontend. We can we can switch between. We can kind of supervise if we want or not. Maybe Maybe while this is happening in the background, I'm gone to the gym. I'm going to go lift. I'm going to run. I'm going to do something. Who knows? You know, and so it's just like searching my emails here. Logged in as in in browser, bro. And it's doing some email stuff here and uh I really hope I have enough time for this, but we'll see. So, it's continuing to search and find things. Um, and so it's now finding details about a dinner, which was yesterday. Uh But now it's on Google Maps. You can kind of see the background update there.

23:20How cool. Um, and it's trying to find the distance between the hotel and the dinner thing. Um Mind you, all of this is happening while I'm here. I'm present. I'm with you. I don't care about that. That's background stuff, yeah? The vision is I'm I could go off and do something. Um, but the data comes to me. Um, there we go. We got the answer. Uh the hotel is located in Espoo at this address. Um, please come, say hi, ask questions at the hotel if you want. Um, anyway, and the speaker dinner is uh somewhere else. It's in the center of Helsinki and it says you should plan to leave. This is what I wanted, but I didn't want to navigate Google Maps with some cookie banner. I didn't want to navigate all these things. I just wanted to ask and get an answer, right? That's what I want.

24:01Tomorrow, there's going to be so many talks uh about model context protocol and MCP and things like that that create experiences like this. I have like 5 minutes left, so I'm going to show you some code of how you can do this uh and then we'll wrap up. We'll do some Q&A, okay? Um, what uh what this what powers this is something called MCP, model context protocol. Um, the browser, the web is actually inherently a client-server relationship. Your browser is a client, sends a request to a server. You go on futurefriend.com. Services, here's a page, here's context. Model context protocol works the same way. You'll hear more about this tomorrow. It is a client-server protocol where a client, in this case a client is like an AI app, ChatGPT, Claude, whatever. A client calls an MCP server, says, "Hey, give me what? Not a web page. Give me context, model context protocol." So, context is usually tools, like functions you can call or resources, text, um or prompt templates, a bunch of stuff.

25:00Client gets context, uses context. I implemented a very basic MCP server for future front end, just to just to play. And I'd love to uh show you that now. So, this is Claude, right? If you're using Claude, you'll see it. And you can see I have no connectors here. But I'm going to add my MCP server. And in 2026, the best way to add an MCP server is still to edit a JSON file in VS Code. Um which is ridiculous. Anyway, um so, this is my MCP server. I'm going to paste it right here. Um and this is my future front end MCP server. So, now, the best way to refresh Claude in 2026 June is to literally quit the app and reopen it. Um wow. Anyway, um and so, now I should have connectors. Future front end, amazing. And so, I can ask uh who's speaking at Future Front End?

25:47And of course, I'm using Opus for this cuz I need to token max very hard. Um like I I need to burn up all the water on Earth, otherwise I'm not doing my job. Um and so, it's contemplating. I I would love it, honestly, if this doesn't work. I bet Claude maybe has an incident or something. This is taking way too long, so let's check um Claude status. Uh is it red? Is it red? Come on. Oh, it's green. God damn it. Why is this taking so long? Let's try it again. Um while it does this, what we'll see is we Yeah, okay, so it wants to get speakers from my MCP server. So, I'll give it permissions and say like, yes, here's the speaker lineup for Future Front End, right? And it used my server. You saw that it asked for um it used the integration. These are all the talks. And it looks kind of nice.

26:35Here's mine at 15:30. Cool. Um Tell me about uh Rachel's talk, right? Something like this. And it will just give me So, I don't have to leave my house and go to some website. I'm in this environment, and the data comes to me. Does this make sense? Um I kind of like it except when whenever Claude's not available. Um, I kind of like it because the only thing that can go wrong here is that Claude becomes unavailable. What what doesn't go wrong is cookie banners and things like that, you know, and so um, ah, it's a real shame it's not. But anyway, I I get the data here in my environment. But if we're talking about taste, right? This is kind of soulless. Would you agree with me? Like, we kind of lose the brand of future front end. Like, they did work. They did work in creating a logo and a color scheme and like, we just lose that, which sucks. Imagine Coca-Cola but without the red, you know, it just it's not Coca-Cola anymore. It's ChatGPT Cola. It's something else. And so, how can we fix this? And this is the last thing I'll show you today.

27:42Um, this this is really so like, let's actually compare. Let's go here. Look at this, beautiful. Light mode, right? Beautiful. You see, this is the work that they did with their branding. There's pictures of the app the speakers and so on. Soulless, right? Um, what can we do and can we do better? The answer is yes, we can. Uh, part of the MCP specification is something called MCP apps, where developers and design teams can create their own little widgets that are embedded inside AI applications like ChatGPT and Claude and others. This is the perfect tasteful middle ground to get the design work of the design people, their taste inside an environment that you know, something like ChatGPT. Let's look at this now and we will finish the talk. Uh, to do that, we'll use my MCP server, but this time slightly differently. So, close this.

28:34Uh, I'll NPM run my MCP server here. And what I'm going to do is this time I'm not using Claude, I'm using ChatGPT. So, I go to chat.openai.com/apps and I will add my own app here. So, I'm going to close this and go to chat.openai.com/apps. And here, um I'm going to add my own app to chat.openai.com. This is my MCP app that I created. So, I'll create an app and I'll call it uh Future Frontend. The description, I'm going to leave it blank. I need a server URL. So, my app is running locally on localhost, but I've tunneled it to the internet. So, this URL actually accesses just my localhost. Thank you, Cloudflare. Um there's no authentication, and I know this is super risky, uh but I I kind of don't care. Uh and I'll I'll create this.

29:19And so and so now what's happening is OpenAI is calling this HTTP URL uh that ultimately resolves to my localhost and and my running MCP server on my computer. Um and it's just going to consume my MCP server and says, "Cool, do you want to add Future Frontend to chat.openai.com?" And I'm like, "Yeah, sure, no problem." And so now it's added. So, now if I go start a conversation with chat.openai.com, right? Uh I'll ask, "Who's speaking?" Uh just like this. Speaking. And the answer is going to be um "Which event are you talking about?" Haha, that's so funny. Uh intentional because I'll say, "Who's speaking at" and then I'll at-mention Future Frontend. I I just made that, right? And now it's going to use my MCP server that I wrote. Check this out. Uh and it's going to Oh, look at that. It's going to embed the speakers with avatars, with cards. I made this. With the logo of Future Frontend. Like it's branded. It's beautiful. It's personal. Um and it's true. There's Rachel Lee, for example.

30:21There's Ramona who just spoke. Look at that. Um it's just tremendous. And and there's also text, so it's accessible. And I could do even better. I was like, "Tell me more about which speaker. Who should I pick on?" Anyone? Huh? Rachel. Okay, Rachel is popular. Tell me more about Rachel's talk. And uh and now I even get to details. It's because I MADE LOOK WOW, WITH A PICTURE AND LIKE THE TALK proposal and which stage is incredible. And there's links. Okay, but are they even interactive? Oh, they're totally interactive. I can just open like external link. It's crazy. So, like Wait, wait. What happened? Is this not right? It's not You changed Well, Rachel changed it. What about X What about this one? This That was kind Okay, there we go. Yeah, that's Rachel Neighbors. So, so it it just it totally works and it's beautiful and it's tasteful, right? In fact, here, let's let's As I'm finishing, this is what you get with MCP apps. This is what you get with MCP. Look at this contrast.

31:21Absolutely night and day, right? And that's how we can bring a little bit of taste um to what otherwise would be pretty soulless AI world without the cookie banners that like the host application ChatGPT. We say yes or no to cookies on ChatGPT. That's it. Their accessibility decisions affect the rest of it. Now, of course, should we consolidate power like that into like one company? Probably not. Good news is you can make your own. You can make your own MCP client and then just render beautiful things. So, let's wrap up. We talked a lot today about taste, about the new front end, what it might look like. Do we even need browsers at all? Maybe, maybe not, right? I hope today I've just given you ingredients for a dish that you will cook. I'm not here to thought lead and prescribe things, but I am here to just show you where things are at and inspire you to build the next few things. In practice, what does this look like? We have to think about taste quite often because we create a lot of software for a lot of very rich companies and and people at IBM. And we put a lot of this into practice building something called Open Rag, which is um Like large Fortune 500 companies, for example, enterprises, they have a lot of data locked up in like Teams calls or like Excel sheets. And we built a system that tastefully just gets operationalized all of it. And so that's a open source thing we work on. For now,

32:43just thank you so much for coming out and for listening to this talk. Thank you. >> [applause]

More talks

All 19 transcripts

Elsewhere

There is every talk I have given, all 69 of them, ConTejas Code, the podcast, and Fluent React, the O'Reilly book on how React works inside.