What's The Framework of the React Future?

[music] [music] We've had a wonderful day of seeing new collaborations for the first time, seeing new integrations into frameworks that we've been using for a very long

time, deeper partnerships through the React Foundation. There's been so much

React Foundation. There's been so much collaboration. It's been amazing. But we

collaboration. It's been amazing. But we

know that you also like to see a little bit of drama. And so we're going to take this collaboration and see if we can dress it up with a little bit of competition.

[laughter] Yes. Here we go. leading a colossal

Yes. Here we go. leading a colossal panel with a size of eight magnificent fabulous framework developers that you

know and love. All chasing that blazing fastness, answering the one question on everyone's mind. Which framework will

everyone's mind. Which framework will ultimately win the framework wars?

Welcome to Framework Fight Night. Please

welcome YOUR FIRST FIGHTER, THE BLUE COLLAR CODER, JACK HARRINGTON.

[screaming] >> HEY, before you leave, before you leave, get everybody to give him a big round of applause for Michael. He's been doing a fantastic job. Come on. Seriously. Yes.

fantastic job. Come on. Seriously. Yes.

[laughter] Absolutely. Good work, Michael. I'm not

Absolutely. Good work, Michael. I'm not

actually a fighter. I am the guy MCing the fight. So, all right. Let's bring

the fight. So, all right. Let's bring

out my my contestants.

All right, Michael, who's going to represent Mikall who's going to represent Rock the Rock framework.

Josh Story for Nex.js.

for Nex.js.

[applause] Tanner Lindsley for Tanstack.

>> [music] [applause] >> Can't see Dods for remix.

[applause and music] Peter for Redwood. You just saw him a couple minutes ago.

[music] [applause] Evan for Expo [music] [applause] and Deon for Adobe.

Come on out. All right, guys. [music]

So, we're going to do do a little bit of formal thing from the beginning and then we're just going to kind of get into it.

So, I'm just going to give each one of you about two minutes minus you Tanner because you just had 10 minutes.

>> Agreed. uh [laughter]

to just kind of give us an intro. Tell

us kind of give us a sales pitch like why should we use your framework? Why is

your framework the framework of the future?

>> Uh yeah. So I'm uh representing rock which is React Native framework. We're

trying to do two things right. So not

all of the things. uh first is getting these uh older React Native community CLI apps uh on the newest tooling so they can leverage uh some local and

remote caching which is especially helpful with picker teams and the second thing is uh getting the native

applications so iOS and uh Android apps into the react native realm. So, uh,

with our rock brownfield feature, we allow them to simply import one file and then just instantiate React Native in

their iOS or Android app. That's

awesome. All right, Josh.

>> Hi, everyone. I'm Josh. I did not come prepared to fight, but [laughter] I I'm on the NextJS team at Verscell.

And >> there you go.

>> Nice. Um,

>> GMO, >> hopefully some [laughter] of you all know that. Um, so Nex has been around

know that. Um, so Nex has been around for a while, so some folks probably are aware of what it is. But, uh, it was started with, uh, you know, principles that were very similar to what React

cares about, which is allowing people to build the best user experiences easily.

Um, back when it began, serverside rendering was sort of not super common and, um, you know, sort of Jamstack SSG was kind of like coming into um, sort of vogue. uh and the framework kind of

vogue. uh and the framework kind of adopted that uh mindset. But you know sort of picking a specific implementation or technology is not super important. What's important is you

super important. What's important is you know can people ship fast? Can they

understand can can new uh coders sort of come into um sort of the community and be productive and focus on what's important like building the product. Uh

and in 2022 started working on a new version of Nex.js that uh we later shipped as like a incremental upgrade called App Router. And hopefully some of you all have heard of that or used it or

tried it. Um, and this is the React

tried it. Um, and this is the React server components uh uh sort of implementation that has been stable for a few years now. Um, I guess to answer the question of like why you might be

excited about using in the future um you know the the journey is not over. Um

there's something that's sort of essential in any React uh project and that's composition. And today there are

that's composition. And today there are still some places where there's some hard cliffs. you have to think about

hard cliffs. you have to think about whether my site's a static site, it's a dynamic site. Um, and so the stuff that

dynamic site. Um, and so the stuff that we're working on now, and actually we have a conference in two weeks if, uh, you'll want to tune in to some announcements. I'm going to not try to

announcements. I'm going to not try to spoil too much of it right now. Um, uh,

we're going to announce some interesting stuff around that and, um, making it so that, uh, sort of these, uh, features that have been in R&D for a while, like partial pre-rendering, uh, we can, you know, sort of share a progress update

there and hopefully bring it to, uh, more users. So, uh, that's my that's my

more users. So, uh, that's my that's my bid. Awesome, Kent.

bid. Awesome, Kent.

>> Oh, yeah. Sorry. Uh, [laughter]

I was just waiting for Tanner's. Uh,

yeah. Um, so I think that the framework of the future is React. [laughter]

Um, and that's what honestly the interesting thing that uh I've observed is that the frameworks because React is doing so much more for us now, the frameworks are really converging and

they really look similar um uh now. So

for for my bit I think um probably the the most compelling reason to use React Router is if you're already using it um which uh a lot of you are uh can you can continue to use it and take advantage of

all the latest features of React uh for those who are looking for a new framework you're starting a new project you cannot make a wrong choice uh these all all of these frameworks are fabulous

uh for React Router in particular um I really appreciate the u the dedication of the team. Uh it has a lot of investment from Shopify. Um I actually am probably unique up here in that like

I've only made two or three commits to the framework that I'm representing.

[laughter] So um I honestly I I'm just here representing the hard work uh from Ryan Florence and Michael Jackson as well as the rest of the team um and all

the the stuff that they have done. So um

I'm just happy to be here and I I feel really really good about where we are uh from a framework perspective now and I just think it's really important uh for frameworks to take care of their users

and I think that React Router has done proven to do that over the last decade.

>> Yeah, absolutely. Peter, uh I'm not going to tell you to use my framework because you're already using it.

[laughter] If you're using V, TypeScript, and React using Redwood. There's nothing to learn.

using Redwood. There's nothing to learn.

You already know what we built. Uh, it's

lightweight. It's composable. It's

server first.

There's no magic, no type generation.

Click command click on a function, you go to where it goes. It's

understandable. It's traceable.

>> All right. There you go. Join the point.

Nice. All right.

>> Evan.

>> Yeah. Um so I work on Expo Expo router uh universal react framework and we really believe in uh just the the beauty and the optimization of the mobile form factor and how much that resonates with

real people and and obviously there's just so many incredible things about the web. The distribution is incredible. We

web. The distribution is incredible. We

just spent 15 minutes stumbling through test flight backstage. Um and like the the web has no such issues. And so like how how can we bring the the best of the

web uh to to these these clearly incredible devices where you have like haptics and gestures and such like thoughtful careful design and uh we were really happy with what we built here.

You starting with React, you had like you know bring JavaScript and React over to native but now we have filebased routing, server actions, API routes, environment variables, one-click

deployments. Uh so there's just so much

deployments. Uh so there's just so much good stuff that the the web community has innovated and uh it's incredible to bring it over to these um these new spaces, the new form factors. So highly

recommend you guys try it. Check it out.

Can build an expo website and get started instantly.

>> Awesome, Devin.

>> Um well, I guess I'm sort of the odd one out here since I don't really represent a framework per se. Like Brussels is just a bundler. Um, so I guess if you have like an existing like client app and you want to like try out server

components for example and you don't really want to migrate anything or like you don't want to add SSR, you just want to like like I showed in my talk like embed some server components into your

app and try it out like a bundler like parcel is an easy way to do that. Um,

also I guess um I I really like what the React Router team is doing actually with um where they have you can actually swap between different bundlers in React

Routers like data mode which is really cool. Um so if you have like bundler

cool. Um so if you have like bundler specific plugins or types types of things like that that you want to use with React Router you can you can actually do that. So um that's super cool. Yeah,

cool. Yeah, >> that's great. So is the framework of the f we have an amazing situation here because we have folks that represent both web and native and is that

framework of the future and and react can do that. Is the framework of the future a framework that spans across web and native and I'm just going to throw it to anybody who just wants to jump in on that.

>> So it it already does.

>> Okay.

>> React.

>> Oh [laughter] that's true. Yeah. If you

think about React as the framework of I mean I think when people think frameworks nowadays they they think about uh the stuff that's on top of it like you know the the routing and you know all the mechanics that are on top

of React to get to where like a full app.

>> Oh meta framework.

>> A meta framework. Okay there you go.

What is the meta framework in the future? And it is is it both React and

future? And it is is it both React and React Native.

>> Yeah. Well the web is just extremely thoughtful about thing. And we see this especially on like the first day in that React keynote and introducing you like the activity API. We have primitives sort of like this on native with like

view controllers and fragments. Um but

there's clearly you know something there and the React team goes and creates a very thoughtful API that just works everywhere. Uh so I think we will

everywhere. Uh so I think we will continue to bind and and sort of meet in the middle here. And it's really exciting because I look at things like activity and the view transitions and it's clear how you can start to build

things that previously you would only see in a native app in the browser and you get that distribution. But at the same time then you have like liquid glass and transitioning through HDR

happening on iOS. So like the bar keeps keeps moving. Uh but we do see these

keeps moving. Uh but we do see these incredible like things transition downstream and overall just you know meets in the middle and it looks fantastic. So

fantastic. So yeah.

>> All right. Anyone else want to pitch in on that? All right. Uh, how about RSC's?

on that? All right. Uh, how about RSC's?

What do you What's the take on RSC's?

>> They're neat.

>> They're neat. Nice. Okay. Are they

required? Like, do you think a frame a a meta framework can't be a meta framework without RSC?

>> I think for React, yeah, probably. Yeah.

One of the beautiful things about React is that it has kind of um eaten up the the uh use cases that we have had in frameworks and and solved some of these problems. And if you don't support that,

then you're only supporting part of the ecosystem. Um which I think I'm speaking

ecosystem. Um which I think I'm speaking from the perspective of a a web developer. But uh yeah, I feel like not

developer. But uh yeah, I feel like not uh supporting that is uh kind of shutting off a a big portion of the ecosystem. And that's I mean is anything

ecosystem. And that's I mean is anything required? you could probably get users

required? you could probably get users without support for RSC's. It's easy for me to say this now that React Router has RSC support. Um but uh I do think that

RSC support. Um but uh I do think that it's as far as framework of the future for React feels like a requirement to me.

>> I think that um obviously like I don't believe in absolute. So there's

certainly people with situations that don't need it or don't benefit from it.

But one dimension that is kind of interesting is that um your users are probably not close to your data. And

being able to have a programming model where you can choose where you want to run certain kinds of compute um unless you can somehow magically make sure there's never serial data dependencies.

having uh you know one query resolve and then lead into a second query if that's right next to your data store that's always going to be better than having to go across the world and do that a few times. And so there's just like

times. And so there's just like structural benefits, not that it's necessary in all cases or you know uh certain apps that are super localized or whatever, but having the ability to make

that choice is is really nice. And then

another thing is that um uh when you constrain like what you can do um it allows frameworks to optimize more. And

so um like it's maybe frustrating that you can't use certain things like I don't know use state in a server component. If you didn't think about it

component. If you didn't think about it too hard you might think like oh why why can't I do it? I I don't want to think about like what does it mean to be a server component versus a client component. But if we did allow that then

component. But if we did allow that then we have to figure out how to do updates over the network and and what's that really serving. So um you know in the

really serving. So um you know in the design of uh server components sort of as part of like the react architecture as a whole um you know you pick your

limitations so that uh the tools that people build with them can be even more powerful and hopefully the choices that you're making are the right ones for the most people in the most situations. And

I think that basically we might all have slightly different takes but largely I think there's a lot of alignment.

>> Yeah. I think something that we haven't seen a lot yet, but I'm looking forward to seeing is like more libraries like just on npm that are server components, right? Like now that more frameworks are

right? Like now that more frameworks are going to be supporting server components, I think this is going to be possible so that you maybe maybe you'll have like authentication libraries or like form validation libraries or these

kind of things to just get componentized now and they can be used in any framework. I think that's what's really

framework. I think that's what's really powerful about server components. It

would be great to see something like shad CN for functionality.

>> Yeah. Right.

>> Oh, yeah. Yeah.

>> Or features.

>> Well, I hate to bring the a word into this, but uh AI, how's that going to fit into [laughter] >> into the framework? Do you do we expect

to see frame meta frameworks have features in it that make writing AI apps easier or do you see AI as being a thing

where it we write frameworks in a way that makes it easier for AI to write them or is or some of the all the above?

>> Oh, for sure it's all the above in my opinion. I think like um at this point

opinion. I think like um at this point if if you're building a framework you really need to be considering how easy the structuring things in such a way

like collocation helps a lot with AI right this is one of the reasons why AI is so good at tailwind um because it's all colloccated right there so like thinking about the way that you design

your API so things can be colloccated amongst other things and then um as far as like making a framework that It makes it easy for you to use AI. I'm not

really sure. I see that as like a framework feature, but I definitely do see like publishing an MPM package that handles both the server and client uh aspects and you just render a component

and boom, now you got app.

>> Yeah, you got it. Yeah.

>> Good to go.

>> Thing like that.

>> More thoughts.

>> Yeah. So, um I think there's obviously u sort of a quality of react apps where you can locally reason is obviously super beneficial. I think that was

super beneficial. I think that was probably mentioned in the keynote too.

Um and I I imagine that all of our frameworks here essentially preserve that because we're all React frameworks.

Um uh and and the less there's like spooky action at a distance, the easier I think it'll be for like tools to write uh uh sort of what you would otherwise write yourself as a software engineer

building with these tools. Um, I do want to call out one thing though that I think is kind of interesting. It it

wasn't that long ago that um you couldn't SSR and stream like that was just I mean you could like obviously stream the bites to the browser but like there was this concept of like

progressively rendering like a React app from the server. Uh and now we have uh sort of plethora of apps where you might you might be streaming for 20 minutes

because there's tokens being spit out of LLM model. Um, and so it was sort of

LLM model. Um, and so it was sort of probably fortuitous that it happened to line up time-wise where the the apps came after the capability did, but um

there's uh that that's carrying a lot of weight and maybe we don't think about it too much because it just was already there ready to go. Um, but it would be much harder to do that at least from the

server um if uh you know we were stuck in the days where uh you know had to buffer everything until you had finished the render and then you sent it to the browser. So,

browser. So, >> so streaming sounds like table stakes for this.

>> It kind of feels like it, >> right? You know, every framework has to

>> right? You know, every framework has to support it.

>> I mean, I won't say every anything. So,

but [laughter] >> uh certainly I think next sport, you know, obviously supports it. I know

Tanstack does.

>> All right. Uh how about AI writing frameworks? I mean, the collocation in

frameworks? I mean, the collocation in that it's obviously a big word, but like having all the code in essentially one file, very close, the AI can kind of parse through it. Are there other things that we can do to help AI write to our

frameworks right for our meta frameworks?

>> Yeah. So I think it's uh if it's easy to for humans it's probably easy for LMS as well right. Um so uh stuff like

well right. Um so uh stuff like composition and uh collocation that works. Uh other than that from the

works. Uh other than that from the framework perspective uh we can write some some helpers toolings. Um there is like plethora of different ways you can

I don't know uh call to an emulator or a device. Uh and uh having a helper

device. Uh and uh having a helper library that you can just expose to LM so it can just call it and call it a day is going to be easier uh for it than

figure out how to do it on you know on its own. It it's going to figure it out.

its own. It it's going to figure it out.

It's going to just uh burn more tokens.

So uh yeah to me it's uh if it's easy for for a developer it's going to be um it's going to be easier for LM and there is also like a level of seniority um to

that. So uh we have junior we have

that. So uh we have junior we have regular seniors developers uh and they're using different tools. So uh um sometimes we may get away with uh

logging our stuff in the console. So our

LLM can just cross reference in the terminal and see uh so we may uh think that we need console login in our frameworks. Uh but then someone figures

frameworks. Uh but then someone figures out MCP to Chrome DevTools and you're like oh now I'm like regular senior

developer uh running this React DevTools and I can read from that.

>> It's funny you bring up MCP because that would be Oh sorry.

>> Oh no. Oh, I was going to say I mean that would be another thing we'd talk about in terms of future facing features, right? Having our apps support

features, right? Having our apps support MCP native apps supporting MCP now Apple's working in that space but also MCP endpoints you know is that something

you see in meta framework something is going to come out of the box what what do we see in that space? Uh next is exploring it. Um you know right now

exploring it. Um you know right now focused on developer experience like uh you know we go through a breaking change. Uh how do you kind of contextual

change. Uh how do you kind of contextual like code mods are great but like at some point there's sometimes situations that can't be code modded or maybe you didn't even know to run it. Um but you can have uh sort of some interface with

the framework to understand like its state or its behavior or um some incompatibility that you might have um either newly introduced or it was there before you made an upgrade to a new

version. Um so that's kind of how the

version. Um so that's kind of how the next team is currently exploring it. I

don't know if there'll be more. I'm sure

there will be more. Um but that's actively being invested in right now.

>> Kent, what's your take?

>> Yeah, I you're all over this.

>> So many things to >> Yeah. Right. Yeah. Um, so in particular,

>> Yeah. Right. Yeah. Um, so in particular, um, like we we talked about, oh, now you can make a chatbot with an RSC component. I I actually think that, um,

component. I I actually think that, um, we're all used to adding chat bots to our applications and now the chatbot is actually adding our application. So

we're get getting the opposite, uh, experience and that's all coming through MCP UI. Uh, currently OpenAI's

MCP UI. Uh, currently OpenAI's implementation has something custom, but they're working together to make that uh, cohesive. And um right now using

uh, cohesive. And um right now using MCPUI like it's possible. I'm actually

flying to Toronto tomorrow to go talk at RemixJam about MCP UI with React Router.

And um that experience is fine, but it could be much improved and I think that there's room for frameworks to support that. Um, this is a a big like bet that

that. Um, this is a a big like bet that I'm making personally on my career, but if it does play out to be the case, then I think that every one of these frameworks will want to have some um

affordances for MCP and like serving your app right into that chat experience. Uh, whether it be chat GBT

experience. Uh, whether it be chat GBT or Gemini or Cloud or whatever um that will be an important part of the future for React frameworks or any uh web framework.

This is bringing back into play the whole micro front end thing. I mean

essentially an MCP endpoint delivers MF back. Yeah,

MF back. Yeah, >> I don't know enough about micro frontend. [laughter]

frontend. [laughter] >> Okay, that's fine.

>> Have an opinion.

>> You know, it's just terminology, right?

It's all the same thing. All right. Any

more ideas on MCP? Are you guys who's who's engaging with MCP in your frameworks?

[laughter] >> Yeah. Okay, fair enough. Yeah, I added

>> Yeah. Okay, fair enough. Yeah, I added an MCP MCP starter to Panzax card. Evan,

you >> Yeah, I mean, we we maintain an app called Expo Go, which is sort of like a a browser for prototyping. And in the same way that you might want like an MCP just built into Chrome to help you

interact with it. We're exploring just doing the same thing with Expo Go. So,

you know, understand like what's on the screen, help you navigate what the current sort of like URL is in memory.

Um, >> so like a debugging tool.

>> Yeah. Yeah. Yeah. Like inter like just integrate directly with it and and and just see like how far you can push the the agentic process. I mean the main thing we found is you just really want to be like careful with your context

window like really sensitive what you'll expose to it. Like image models can you know essentially poison it in like multiple different directions to be like is does the thing render and it'll be

like start just describing like the status bar and then it'll like get confused. Little little pieces like

confused. Little little pieces like that. Um the

that. Um the >> I know last year your talk was one of the highlights here.

>> It really around generative UI. It was

fascinating. So is that something that you see being a standard part of Expo?

Where where did where's that going?

>> Yeah. Um I mean like kind of to Josh's point like streaming wasn't big. Now

streaming is just built into everything.

So, we have all these streaming primitives and we found that that's just been huge for a number of like the the Mistral app is an Expo app. Um, I think it's hard to name an AI app right now. I

mean, I guess it's not that hard. Uh,

name that isn't like fully expo, but uh like even the top dev tools like in the top 10, I think four out of 10 are are expo tools and a lot of them are just helping you build expo apps. So more so

than like expo adding AI, uh we we want to be a really good tool for helping others build AI tools, right? Like very

recursively because when I you know I just moved to San Francisco when I talk to people no matter who it is, everyone is trying to build something agentic and so you know downstream of that they're

going to reach for a dev tool like Expo.

it needs to even something as simple as like the the masked shimmering text effect where it's like thinking like that was slightly harder than it needed to be and we found that that was like

you know everyone was stubbing their toe on it. Um so yeah we're helping like in

on it. Um so yeah we're helping like in basically any way we can to make it just a machine that builds machines. Okay,

agents seem kind of murky. Like, is

there anyone who wants to like kind of talk about solidifying for folks what agents mean in terms of like us developing agents for customers on these meta frameworks?

How's that going to work?

>> I nominate Ken to describe that.

>> Yeah. Uh, so you're asking for a definition of agent.

>> Well, I you know, just like when Well, obviously folks think about like the Yeah. the product and all that like you

Yeah. the product and all that like you know >> it going off and doing things on its own is that something that like on running jobs and tasks is that going to be >> a critical part of a meta framework that

just something the backend's going to >> uh I I certainly it's going to be a backend uh thing. Uh here's a way that I I see that going. Um

>> let's say that you're a bank and you're considering like opening up your capabilities to an agent. That's

terrifying, right? uh [laughter]

um an agent that you don't control like okay we're going to make an MCP server for transactions and now like you have no control over that experience. So um

what what if instead like having an agent that you build yourself you get to control that experience like yeah maybe that's okay I get to manage that and so what you do is you make an agent and then you turn that into an MCP server

and you expose a chat tool and now um claude or chatbt like I set up your MCP server and my agent talks to your agent that's literally I'll have your people

call my people or whatever right like um and and then let's also like same sort of idea you've got a really big organization with a lot of different things that it it can do. You're going

to have little sub agents for each parts of those uh pieces of the organization and each one of these can be um MCP server that communicates with all of the uh different pieces and then just

exposes a mechanism for communicating across those effectively a chat tool. So

for us frameworks uh as as far as react is concerned I I think that there are elements of like well okay the bank is even though it's a chat tool maybe there is some mechanism for sharing some UI

think that makes sense or like you're you know uh showing a map or something like that your Uber um and so there like we're we're still in that realm but I think a lot of that agent stuff is going

to be on the back end and whether or not react frameworks kind of swallow that piece uh I think is I I wouldn't say it's required certainly, but I think it could be useful and frameworks like to

do useful things. So

>> yeah, I mean I think you know as we saw RSC kind of expand the scope and and nature of React, right? Seems like

there's a potential there, but it seems like more like a meta framework type deal.

>> I think maybe there's more for the native folks um on that like ondevice uh um LLM like that's the future that I want. Um, so yeah, get to work, guys.

want. Um, so yeah, get to work, guys.

Thanks.

>> Uh, I mean, Apple's ondevice foundation model. Do you you want to guess the

model. Do you you want to guess the context window size?

>> Very small. I imagine [laughter] >> it's like 4,000 characters.

>> Yeah. Well, you can make lots of sub aents. [laughter]

aents. [laughter] >> Yeah, >> it'll get better. That That's my expectation is that it will improve.

>> Yeah. Although this is like 4K is today.

Like let's imagine two, three years from that. And there's going to be like most

that. And there's going to be like most likely there's going to be uh more progress here. Plus uh nice thing about

progress here. Plus uh nice thing about uh the local is that it's not that you only have to use the the one provided by

Expo by by Apple or Google. Um, but you can just download your own um the the one

that's uh that that you like to use uh use MLC or Exeutor um to to run these models. Uh so you're not like locked into this uh um this

particular choice that uh Apple made, but it's very efficient in terms of uh um for as many apps as possible to use the same resources. So you as an app

developer you don't really have to think about how many uh how much memory you're uh using your app is using uh and so on and so forth. Um so I think there's uh

there's some some pretty interesting future for the local OM um also in in in terms of generating the um the UI that

uh like hyper personal UI. Um so right now it would uh make uh likely more sense to um handle that through the

server. Uh but with more powerful device

server. Uh but with more powerful device uh we can have like millions of uh versions of our own UI and yeah good luck testing that >> mining bitcoin or whatever.

>> But it's uh it's going to happen being crypto ads.

>> Nice.

>> Uh all right so we have a lot of members of the react team here. So

what are your meta framework authors and contributors? What do you want to see

contributors? What do you want to see from React? Are there things that you

from React? Are there things that you want to see React take on?

>> Always wouldn't touch that one.

>> You did it. You did it. Good job.

>> Yeah. [laughter]

>> Yes. All right. Don't change suspense again, please. Actually,

again, please. Actually, uh speaking of changes, one thing when we were talking about um making things easier for AI, um not changing things makes things easier for AI. So that's a

big challenge for frameworks as well is like how do I make it actually Theo said this yesterday, how do you make it so that you can make changes to the how it works without changing the API? So you

can continue to improve it but um not change things.

>> Well, I think we're seeing some experimentation in shipping vibe rules.

In fact, they the packages call vibe rules with the packages as versioned resources. So, you know, as a framework

resources. So, you know, as a framework author, you can actually put in, okay, as we generate our docs, we're going to generate rules for LLMs and package it up. So, it's got it's in that version.

up. So, it's got it's in that version.

>> So, it's definitely evolving. So, seeing

some stuff in that space.

>> All right.

>> Yeah. Then you're asking your LLM to actually read the docs and the [laughter] not >> Well, sometimes they do a better job at it than we do.

>> Yeah. I guess they they're getting lazy, right? So

right? So >> yeah, >> no features for the reactive at all.

>> Is this is this like AIcentric because that's kind of where we went, right? Or

are we just back to just like what do I want out of React?

>> Well, yeah. I decided like if there's folks that are like AI, I'm done. You

know, I've seen a lot of that this conference, you know, let's let's talk about React and and K was talking a lot about React. So yeah, let's talk about

about React. So yeah, let's talk about React. What are some stuff are there

React. What are some stuff are there futures in React that you think are missing or that would be really cool to see you would like would make your lives as framework meta framework authors

easier? Yeah. Um I would love to see

easier? Yeah. Um I would love to see >> or you can go and then I'll go.

>> All right. It's not AI really. No

worries. [laughter] Uh

I would love to see uh more like burring the gap between the React and React Native. So what uh uh what the teams uh

Native. So what uh uh what the teams uh showed today with more u web APIs uh uh

better access to uh to DOM and uh um like this is making it easier for web developers to actually be productive in

the uh in the mobile realm. So any uh effort in this area uh is like very welcome from from my point of view. Uh

so I would love to have more um web APIs, more like better CSS support.

Great to see what uh Expo is exploring with uh CSS for for React Native. I

would love to have uh to be able to use the same animations uh and use view transitions uh in React Native. Right

now it's not really possible. It's

probably not going to be for the next couple of months uh optimistically. Uh

so yeah that's uh my wish list. Cool.

Tammy, >> uh, I saw a lot of talk this year about like performance and, um, increasing performance of React, which is nice to

hear at React Comp. And um that brought to the top of my mind like uh as a as a framework author that's supporting not just React but multiple frameworks um I

have to deal with React wanting to own more and more of my state so that I can use these awesome features like transitions and suspense and view

transitions. And um part of that part of

transitions. And um part of that part of that discomfort there is the lack of um something like a use external store. Uh,

I know the React Core team is probably going, "Oh, Tanner's asking for this again." You know, um, but it it really

again." You know, um, but it it really would uh allow me and the tools that my team builds to be able to take full advantage of all these awesome features

of React. Um, but also in a way that we

of React. Um, but also in a way that we could continue to own our data and own our architectures out outside of React.

Um, that [clears throat] that would be very cool. Uh, and I know that they're

very cool. Uh, and I know that they're aware of that, but saying it out loud is, you know, [laughter] cathartic. So,

cathartic. So, >> nice. All right. Yeah. Okay. So, uh, no

>> nice. All right. Yeah. Okay. So, uh, no one. Yep. 10.

one. Yep. 10.

>> Can I just say one thing I think is interesting that none of us said signals.

>> Oh, yeah. Absolutely.

>> Yeah. You got a Well, >> we're not done yet. [laughter]

>> Is that true? Nobody You Are you all fine with the render render cycle as is?

Uh I would I'm not going to ask for signals but uh kind of in that vein is just more fine grain performance. I know

that we have React compiler and that's amazing. Um and everybody should be

amazing. Um and everybody should be working to like enable that by default and as framework authors we should be working to enable it by default. Good

job Evan. Um I'm working towards that but but also I I really do like uh expressive APIs that I can be deliberate about that kind of stuff. Um, I've been

like dreaming and salivating for a like a a deep context selector type API for a long time. Um, and if that was something

long time. Um, and if that was something that was native to React, again, I I feel like um it would offer a lot of control that I I've missed and and

wanted, which is kind of there with signals, but I'm not asking for signals.

Yeah, you still get the render cycle, but you get that fine grain granularity to pick off the data that >> makes sense to render on. Yeah.

>> And in fairness, we we kind of simulate that in a lot of the TANS stack tools.

We we allow for fine grain selectors, but we can only do that because we use you sync external store. Um, which means we can't support completely the full like gamut of transitions and things

like that. So, I don't I would like to

like that. So, I don't I would like to not be at odds anymore with like fine grain rendering support versus um you know, full transition support. Uh I

would love for just to be able to embrace all of it.

>> All right. Uh any closing thoughts? Got

kind of running late on time. Although,

you know, you jump in on how is Remix V3 going to just shake the apple card?

>> I have no idea what to expect on Friday.

Um but you all should tune in. I'm sure

it'll be entertaining. [laughter]

I'm pretty sure they're putting >> SpaceX launch some stuff in there just to get people upset, I think. [laughter]

>> Yeah, possibly. Possibly. Any any final thoughts for anyone >> on the future of React?

>> I I think it's worth punctuating one more time that like there is just a lot of convergence.

>> Um, and that's a good sign for React. It

means that the stewardship over React is listening to us, listening to its users and and seeing the common threads. Um,

and I think that's the ultimate future we would all dream for is that there is no meta framework that React is just that's the tool that you need.

>> And I mean, that might be a pipe dream, but I think it's a good goal and and something to to keep in mind.

>> Do we think the foundation is going to help with that?

>> I think so. I hope so. Yeah.

>> Yeah.

>> All right.

>> Yeah. Put React in the browser.

>> Okay. [laughter]

>> All right. Can we get a big hand for all of our uh panelists? Thank you guys.

[applause] [music]