Cloudflare Pages with Greg Brimble
===

Paul: ~Test, test, test.~

Greg: ~Test, test, test.~

Paul: ~Okay. ~[00:00:00] Hi there and welcome to Pod Rocket, a podcast brought to you by Log Rocket. Log Rocket helps software teams improve user experience with session replay. Ever tracking and product analytics. Try for free@logrocket.com today. My name is Paul and joined with us is Greg Rumble. Greg is a software engineer over at CloudFlare.

He's working on the CloudFlare Page's team over in the emerging technologies and incubation department.~ So Greg, you studied at the University, university of Edinburgh. Are you calling in from the~

Greg: ~did. Yeah. Uh, I, um, I studied in, brought up in Scotland and I'm now down in London, like you say, working for CloudFlare on uh, CloudFlare pages. And, uh, later this year I'm actually moving to New York.~

Paul: ~Oh, so you've been around, you're coming to the other side of the, the pond. This time.~

Greg: ~I am, yeah, very excited. In fact, today I was, I, um, was me just submitting my Visa application, so I got all that legal and tax stuff to look forward to. But, uh, yeah, hopefully it's all, uh, gonna be coming together and we'll be moving in October.~

Paul: ~Yeah, it's a big main character day for you. You're applying for your Visa and going on a podcast, so well, today~ we're gonna be talking about a whole myriad of topics. There's been a lot going on over the CloudFlare world. Of course, we're gonna be talking about pages because Greg's coming to give us. ~The, the,~ the what we want to know about the latest updates on CloudFlare pages, but we're also gonna be getting into everything that got announced over the week.

~Um,~ so we're gonna be talking about ai, we're gonna be talking about smart placement, unifying workers in pages, web apps, quick edit it, Wrangler three. There's a whole bunch of stuff coming out. ~So, I mean, ~let's just, let's just pick one off the pile, Greg. ~If,~ if we could modernizing the toolbox, there's a whole bunch of updates and that kind of come under this umbrella,[00:01:00] 

Greg: ~Yeah, so~ last week was, ~um,~ our developer week 2023, and we have these sort of innovation weeks, we call them, ~um,~ throughout the year ~and,~ and try and group together a bunch of announcements that ~sort of. Of~ make sense under a single concept. ~So, um, ~yeah. Last week is one of the big ones for certainly my team.

~Um,~ obviously being focused ~for, uh,~ towards developers. A lot of the features and things that we release gets to come under this umbrella of developer week. ~Um,~ and yeah, so ~the,~ the pages team. We look after this product, ~um,~ which is mostly just about giving people an easy on-ramp into the CloudFlare developer platform for when they want to deploy full stack web apps.

~Um,~ and ~um,~ part of what that~ is,~ is a, ~um,~ GI integration. So we have integrations with GitHub. We GitLab, ~um,~ and we will. Look for any code changes that you're making and stuff, and then we will deploy that to our network for you. So if I'm wanting to, ~um,~ deploy greg bramble.com, I have a repo and GitHub make some change, we pick up those changes, build your site, and then deploy it.

So [00:02:00] the whole idea is that's a very seamless, ~um,~ experience. It's supposed to be very slick. But one of the, ~um,~ problems that we were facing was we had this, ~uh, sort of~ antiquated and very monolithic build environment. So one of the announcements we were making last week was this modernizing ~this,~ this toolbox, this build system that we were using.

~Um,~ so there's a bunch of user facing stuff to that. Like we've updated just a bunch of the defaults and the tools that come inside that environment. All the classic stuff you would expect, like node and python being, ~um,~ updated to the latest versions. ~Uh,~ but then we've also done a whole bunch of work behind the scenes,~ um,~ and really broke up this monolithic thing so that hopefully it's easier to keep on top of.

So we continue to update this in the future, but also to set the stage for some of the more exciting announcements, which, ~um,~ we also, ~uh,~ at least talked about. ~Um, Uh,~ in developer week. ~So, um, yeah, I mean,~ we can absolutely dig into some of the technical, if you're interested in this particular, ~um,~ blog post I wrote about this, ~um,~ the build image and all the behind the scenes stuff.

~Um,~ or we can move on to the next one, wherever you like.

Paul: ~Well, well,~ Greg, I'm. Selfishly [00:03:00] interested in one peculiar thing because ~the,~ the node upgrade that you're giving ~to the,~ to the pages environment, it's profound. And we're talking about JAMstack here just to, ~uh,~ paint the picture for everybody listening. These are JAMstack deployments typically. Gotcha. ~So, um,~ you're going from node, what was it, 12 before?

Greg: Yeah, so this is just the, ~um,~ the build environment that we give you.~ So, um,~ let's say you had ~like ~a Gatsby website or something, and when you were running Gatsby build, it's obviously looking, it's using a bunch of node APIs to ~like~ read all the source files that you had and then produce what is essentially just the static assets, the JAMstack part of your site.

~Um, so. Uh, yeah, we,~ we launched CloudFlare pages, I think just over two years ago now. And at that time we were just using like tools, like when they were, when they came out. So it was like node 12 and Python two. ~Um,~ and we hadn't updated any of those. ~Um, um,~ Until now. So people were ~kind of~ stuck on some of the older versions and there was, ~uh,~ obviously a lot of pain there.

If you were just deploying a brand new project, sometimes you needed some of those latest tools and it was [00:04:00] either a hassle to configure yourself or, ~um,~ straight up impossible in a couple of situations. ~So, um,~ Like I say, ~we,~ we ~sort of~ revamped this environment and there's now a V2 beta that you can opt into.

We will be rolling that out to general availability, ~um,~ pretty shortly. ~Um,~ and yeah, it comes with these new versions, so hopefully more projects just work out of the box if you just found something you like on GitHub or whatever. ~Um,~ and it's just a good foundation for people building projects today.

Paul: Because we went from, ~uh,~ node 12 to ~what, what the,~ the current version 18 right now. And so there's a whole slew of other just outta the box APIs that are available to folks. ~Do,~ do you have to do any considerations being like a platform and infrastructure team thinking about how are my users gonna be? are they gonna be building and running now that they can do this? Like for example, fetch like ~that,~ that's in the in node now, is that a consideration that you have to go over with the team and think about, or is it just ~kind of~ like it is what it is and it's gonna be fine?

Greg: For the most part, it's relatively simple. ~Um,~ these are just the defaults. So ~like, um,~ they have been [00:05:00] generally configurable for people if they needed like 16 before or whatever, they could set that. ~Um,~ the jump two 18 was ~a,~ a bit of a stretch just because there were some underlying operating system stuff that had to happen.

~Um,~ but ~um,~ now that we've done that, ~um,~ We're in a pretty good place and like I say, we broke it up. ~So, um,~ there's even more isolation between ~like~ the system tooling that runs in this place, ~uh, in,~ in our build environment, ~uh,~ and the user space. So that will allow us to, ~um,~ let you do whatever it is you want inside this build system.

~Um,~ knowing that you're in a protected and isolated environment without interfering with anything that we're doing on our side. ~Um,~ And so hopefully you'll get even more control as to what you want to do ~in your,~ in your build environment. ~Um,~ It opens the door for a lot of things in the future. Like one thing we've even been thinking about is ~like, um,~ not only perhaps providing you with like newer versions, V three, V four, further down the line, but e maybe even, ~um,~ offering users the ability to just brilliant.

Bring their own docker image and they can just run whatever it is they want in their build. ~Um,~ so that's only gonna be possible with this new architecture that we've been building. ~Um,~ I'm not committing to that yet, but~ it's,~ it's the sort of thing that we're [00:06:00] thinking about. ~Um,~ As well as just the additional features that this sort of breakup gives us.

So we can now like cash on either side of this image. ~Um,~ and so yeah, we're starting to build that out and hopefully that ~will~ will be,~ um,~ available to people

Paul: That's exciting news, Greg, that, ~I mean,~ you're gonna give, ~um,~ Google Cloud platform a run for their money. ~If,~ if people can just spin up. Arbitrary Docker containers like that's.

Greg: ~Yeah, well,~ we'll see. ~Um,~ like this is, ~um,~ all the build level stuff. So at runtime we're still using workers, we're still using, ~um,~ v8 and, ~um,~ that has constraints because we want to be able to run this, ~uh,~ with just ridiculous scale. ~Um,~ so you're not getting all of these. ~Um,~ container environments at the edge.

This is just the one-off build that happens when we like detect a gi, ~um,~ event. But, ~uh,~ for sure you're gonna get a lot more power in that build system

Paul: ~I mean,~ you could even support Build X and. Throw in your own sauce if you can. If people can bring in their own build pipeline

Greg: for sure. ~Yeah, there's,~ there's a lot that's possible. So we're gonna need to, ~uh, uh, well, ~yeah. Excited to explore in the future.[00:07:00] 

Paul: ~I mean,~ if people wanted to run their own processes, their own business logic, another product people reach for common is CloudFlare Workers. I'm a personal fan myself, like it's really easy to spend things up, spend things down, and I know there's a lot that came out during the week related specifically to workers.

~Um,~ but there's something. That we might as well use to transition now about ~like~ unifying workers in pages and how they're now almost a synergetic piece of technologies. ~Um,~ so ~how are,~ how are workers and pages helping each other now in a way that maybe they were ~kind of~ siloed before?

Greg: Yeah, that, that's a great way of putting it. ~Um, ~they were siloed before, so workers came around first. I forget exactly when that, ~um,~ was released, maybe, I don't know, 20 16, 17 thereabouts. ~Um,~ and pages, like I say, is only two years. ~Um,~ it's been available. So we ~sort of~ developed these pieces independently.

~Um,~ and pages was really focused on JAMstack to start with, and workers really was really focused on like compute at the edge. But we've realized that with a lot of the modern full-stack frameworks that's out in the ecosystem now, ~um,~ [00:08:00] things like remix, quick Spel kit. ~Um,~ The lines between ~like~ what a JAMstack app was and what, like compute at the edge means have become so blurred that people were able to, ~um,~ use workers and deploy full stack apps that rendered at the edge.

So you were able to get a really dynamic app, ~um,~ that was really very capable, ~um,~ and put it on our platform. Which was just awesome to see. ~Um,~ so we built pages, functions, and I think that was maybe a year and a half that's been out now, thereabouts. ~Um,~ and ~um,~ that was just a way to deploy a worker with your Pages app.

But obviously now we're in a place where pages and workers are, have very similar capabilities, ~so, um,~ we want to unify them. Bring them together so that, ~uh,~ there's less confusion so that a user doesn't need to answer the question of which product should I use to get started when I wanna deploy something to CloudFlare.

~Um,~ and we're starting with just this sort of unification of our ui, just so that your workers and pages are listed all in one place, but we're going to be [00:09:00] bringing a much more comprehensive, ~um,~ unification behind the scenes so that ultimately they will just be like, An application and that's ~what you deploy to pages, uh, sorry,~ what you deploy to CloudFlare.

~Um,~ and yeah, I think it's gonna really open the doors. It's gonna bring, ~um,~ features that are currently missing in pages. To like ~the,~ the projects that are, that we have, ~uh,~ today and workers is gonna get things like the static asset hosting that pages has. So it's gonna be really powerful for developers and we're ~really,~ really excited to, ~um,~ get started with this.

It's gonna be a long process, but it's going to be worth it ~in the,~ in the long term.

Paul: Out of that long process. ~Where,~ where are we right now?

Greg: So internally we've completed stage zero, which was the UI

Paul: great.

Greg: ~So, um,~ I think it's gonna be like multi quarter effort. I don't think this is gonna be, ~uh,~ close to done, even like at the start of 2024, but, we'll, ~um,~ A lot of the team are going to Austin next week, ~um,~ just for ~like~ a team offsite. And we're gonna be, ~uh,~ thinking about this pretty much nonstop during that time, doing lots of whiteboarding sessions and stuff and really figuring out what the concepts are, what all the features are, [00:10:00] where they make sense in this sort of ~information,~ information architecture.

~Um,~ so looking forward to that week and figuring this all out. But we're, ~um,~ obviously already have some ideas and, ~um,~ we'll just be implementing them over ~the,~ the months to come.

Paul: ~So is Smart Placement one of those ideas that you guys have already been working on? Because if,~ before we go into it, what is Smart Placement? We're talking about placing things smartly close to each other. So what are we placing?

Greg: ~So, um,~ like I say, ~uh,~ people have been deploying really capable full stack apps to CloudFlare and, ~um,~ to date they've been running at the edge. But we realize, although that's really great for a lot of apps because you're able to quickly respond to users, ~um,~ they only need to connect to their nearest.

~Um,~ data center to be able to, ~um,~ be able to render the page. It doesn't work for absolutely every type of app. If you've got a reliance on some third party that isn't as global as CloudFlare, let's say you've got a database in the States and you've got visitors in London.

Paul: It's a classic example, your database. Yeah.

Greg: ~Um,~ if you've got ~a,~ a database in the States, any visitors in London, ~um,~ If you need to have ~like~ a noisy conversation with your database in order to render your page.

Let's say you've got like [00:11:00] your user's table and a list of all of their, I don't know, events, ~um,~ that they're attending, something like that. And you maybe need to join like multiple times across the table or whatever. ~Um,~ In order to do that, you would, ~uh, like~ if you were rendering in London, you would need to have multiple requests back and forth over the ocean, ~uh,~ with your database.

And so you're obviously paying a big penalty for that. What Smart Placement does is it moves your app to where we detect the lowest latency, ~um,~ location to be. In this case, we would move it to, ~uh,~ next year database in ~uh,~ America. And then we would render the app there and deliver the final result to the user.

So that means that you only incur that penalty of going across the ocean once, which is obviously much improved. ~Um,~ and smart placement is exactly that, one of those features that we, ~um,~ that makes sense in a unified workers and pages, ~um,~ land. So it's available today for workers and pages. You can go and turn it on, ~um,~ and.

It's a pretty powerful, like I say, if you've got a localized database or any sort of, ~um, um,~ dependence [00:12:00] on, ~uh,~ some other service, like maybe, I don't know, some, ~uh,~ API you're integrating with or something like that, ~like, uh,~ you can't control where they run. So ~we,~ we will move to wherever they are just to optimize it for you.

And like I say, it's smart. ~Um,~ so you just click a button and we'll automatically detect it and move it for you.

Paul: Does it keep track of ~sort of like~ a history I would assume of ~like,~ here's where I've been, here's how it went.

Greg: ~So, um,~ it's ~uh,~ a heuristic thing. So we look at the connections that your worker is making and then see where they're, ~um,~ going to and over time we're going to like, try it out in a bunch of different places and then realize, oh yeah, this is the optimal location, so this is where we're going to move it, and we're only going to make that move if, ~um,~ We see an actual benefit coming from it.

~So, um,~ we are doing a bunch of stuff behind the scenes to work it out, but then also we're going to show you, ~um,~ in analytics in the dashboard. I don't think it's ready quite yet, but, ~um,~ it will be coming. ~Um,~ so that you can see the effect that this is having and like why we chose to make that decision ~and,~ and where it's now, ~um,~ running instead.

Paul: It seems like CloudFlare is, ~um,~ really gearing itself [00:13:00] for the full stack experience here. Like your products are feeling more blended and synergetic than they really ever have, than before. ~Like~ if you can. Use these and then, ~uh, uh,~ CloudFlare pages and workers ~and then,~ and then move them around. I could make a whole web app or a whole application.

~Um,~ yeah. So you guys are definitely moving in that direction, you'd say?

Greg: Yeah. ~Um, I mean,~ over the past few years, the focus has been on developing all these, ~um, um,~ fundamental pieces. Like we figured out compute, we figured out asset serving, we figured out, ~um,~ I don't know, like ~all, all the,~ all the various products, ~uh,~ like object storage, ~um,~ D one databases. So we figured out all these

Paul: Hmm.

Greg: and launched all these products kind.

I mean under the umbrella of the developer platform, but ~kind of~ in isolation. ~Um,~ and now that we have all of the building blocks, we're in a place of ~like,~ okay, yeah. How can we make it so that developers are very productive when deploying to our platform, and so that they can just bring together all these pieces to deploy one app.

Paul: And would you say that you're Yeah, we're really making Cloud f [00:14:00] CloudFlare well suited to make a web app today. ~Like~ if somebody wanted to come deploy a full stack, like you mentioned D one, that's another great example. ~Like that's a,~ that's a distributed database. So ~we,~ we used like ~the,~ the central.

Database in the United States as an example, you have to go across the ocean and call Homer user data, but you could use D one to~ sort of ~distribute that, and that's something CloudFlare just offers out of the box.

Greg: Yeah, so ~we,~ we have, ~um,~ multiple different data type, ~uh,~ data stores. So we have, ~um,~ D one durable objects. ~Uh,~ KV and R two are the four ones that we offer today. ~Um,~ R two s are object store, kvs are key value store, ~um,~ durable objects are. Very difficult to describe. They're like a point of synchronicity. ~Um, uh,~ so that you're able to, for example, like manage, if you were selling tickets to an event, you could guarantee that there's only ever like one sale of that particular ticket, so you don't double book yourself or whatever.

And then D one is our SQL database. ~Um,~ D one at the moment has single primaries, but what we're working on is, ~um,~ replication for, ~um,~ making it. Fa faster reads available, ~um,~ [00:15:00] more globally. So D one is still, I think it's just gone beta, ~um,~ with ~the,~ the developer week. ~Um,~ and with that there was a bunch of improvements made to really speed up the underlying like engine.

~Um,~ but I think Reid Replicas is next on the list for ~like,~ like we say, like part of these global primitives that you really need when building, ~uh,~ really powerful apps.

Paul: Now before we double click on one of the topics we just talked about, ~uh,~ at a surface level, just wanna remind our listeners, for anybody out there who's a builder themselves, if you're building full stack applications, that this podcast is brought to you by Log Rocket. And Log Rocket will help you build your application faster in, in a more reliable way than you were before.

With session replay, error tracking and product analytics, you can find problems and surface them. Faster than ever spending less time debugging and more time building a great app. So Greg, I would love to double click on one of the things we talked about before, cuz we talked about the, ~you know,~ we're upgrading the toolbox with [00:16:00] pages, we're bringing workers into the picture.

You guys are coming with D one, ~uh,~ within CloudFlare workers specifically, cuz that's ~kind of,~ that product's been out for a little bit. I'm sure a lot of people listening have experience with it. A lot of people have. The experience of trying to edit things in the browser and maybe, or whether you're tooling with it with Wrangler, ~right,~ on your home terminal.

But if you're in the browser and you're trying to edit things, there's some updates that have been made and got announced in during the week, right?

Greg: Yeah.

Paul: ~Quick edit.~

Greg: ~yeah. Um,~ when workers first launched, ~um,~ they were. Really quite simple. Like typically you would only ever use them as like a, ~um,~ way to intercept requests and then transform them in some sort of way before reaching your origin. So maybe you like add header or ~um,~ do like a small bit of user off or something like that.

~Um,~ and we had a, ~um, sort of~ basic editor in the dashboard to help you do that and just get started with it. ~Um,~ but. Over time, people have been building more and more complex apps, obviously.~ Like~ that's what exactly what we've been trying to

Paul: It's the goal.

Greg: Absolutely. ~So, um,~ people have been deploying workers made up of [00:17:00] hundreds and hundreds of files, ~um,~ that are written in type script, all sorts of stuff.

~So, um,~ what we've done is we've just refreshed ~the,~ the dashboard. ~Um,~ we've built, ~uh,~ Visual Studio Code in, and you are now in a much more ~sort of~ familiar I D E when you're browsing and interacting with ~these~ these workers, no matter how complex they are. So yeah, a really powerful upgrade to the sort of dashboard experience for editing workers.

Paul: How close is it to VS. Code? Is there like a file explorer?

Greg: Yeah, ~it's, it's the,~ it's the same visual studio code, ~so, um,~ if you're aware of, ~um,~ I think they call them workspaces, GitHub workspaces, ~um,~ and other sort of similar, ~um,~ online experiences, you're actually able to compile visual studio code for the web into just like straight up HTML elements and some CSS and stuff.

And, ~um,~ We've ~sort of~ augmented it with a couple of, ~uh,~ behind the scenes things to ~sort of~ make it really work well with workers. ~Um,~ and we have further plans to make it even better. But, ~uh,~ yeah, ~it's,~ it's pretty much the exact same experience you'll be used to when, ~um,~ using VS. Code locally.

Paul: If people didn't [00:18:00] want to do the visual edit, I know I'm a fan of CLI tools. Wrangler has always been there letting me test. My workers locally Wrangler Dev and emulate my cloud deployment. There's big updates coming at Wrangler. I remember the upgrade from Wrangler one to Wrangler two. There were a lot of things happening.

~They, I mean, I had to,~ I had to change some things, but it was like all great stuff. So what's coming with Wrangler three now?

Greg: ~So, um,~ yeah, like you say, Wrangler one to Wrangler two was a complete rewrite of Wrangler. It was rewritten in type script and, ~um,~ came in with a slew of new features. ~Um,~ Wrangler three, there's only really one significant change that's been made. But that changes the development experience to now use Worker D.

So we released Worker D I think towards the end of last year. ~Um,~ and that is our open source CloudFlare Workers runtime. So you're

Paul: No way.

Greg: yeah, you're able to, ~um,~ download~ and,~ and run CloudFlare workers on your own machine. And so Wrangler now ships with that runtime and that gives you a, ~um, ~Like for ~like, um, sort of~ bug reproducibility with what [00:19:00] happens locally versus what happens when you deploy your worker to Cloudflare's network.

And obviously that's practically invaluable. ~Um,~ if you were, ~um,~ using Wrangler two before, you may have encountered some situations where. Production and your local environment ~were~ were different and they behaved differently, and that was just because the Wrangler two version was using node, whereas Wrangler three is now using worker D under the hood.

So much more accurate representation of what happens on production.

Paul: So what is gonna happen to mini flair?

Greg: So mini flare is still used under the hood mini flare wraps, worker D, so actually it goes Wrangler, then mini flare, then worker D. ~Um,~ because mini flare is still needed for some of the, ~uh,~ parts of our network that we haven't open sourced. ~So,~ for example, ~um,~ KV or R two s are data stores. ~Um,~ those implementations are still proprietary, ~uh,~ but Mini flare will emulate those pieces while still calling out to Worker D for the actual like JavaScript execution.

Paul: ~Gotcha. Very,~ very neat. So the, I [00:20:00] just unifying that developer experience is gonna be beautiful, ~um,~ especially for new devs coming into the Wrangler experience, cuz sometimes there's a little bit of like, where did this environment variable come from and how do I emulate this locally? So ~that's,~ that's phenomenal.

I'd love to hear it.

Greg: Yeah, it should be like, the whole vibe of last week was just really polishing the experience and bringing everything together onto just like a good, ~um,~ hopefully workflow for people, creating new apps, developing ones they've already got, ~and,~ and publishing them to production.

Paul: And I know you guys have also been talking about new stuff and experimental stuff in the L L M AI world. ~Um,~ a lot of companies have been trying to figure out how to leverage this technology to teach accelerate and form, do a million other things. ~Um,~ so one of the things you guys have coming out, constellation seems pretty interesting.

Greg: ~So, um,~ you're right, like AI ~kind of~ took the whole web ecosystem, ~uh,~ by storm over the last couple of months and, ~um,~ CloudFlare don't like to miss out on things like that. ~So, um,~ one of the, ~um,~ Primitives that we're [00:21:00] adding to our developer platform is Constellation. And Constellation is a new product that lets you, ~um,~ run the, ~um,~ machine learning models on our network.

So you're able to upload ~a, a,~ a trained model and then do like classification or whatever using that model inside workers. And it's just another primitive that we've added to our developer platform. Another thing that you can use to make really cool apps. ~Um,~ and yeah, constellation I think is, ~um,~ I'm not sure if it's actually open to people to join.

No, there's ~a,~ a wait list for beta, but you can sign up for that and, ~um,~ get notified when it's ready. ~Uh,~ I've seen some demos of it running. It's pretty damn cool. And there's some examples in our blog of exactly what that's gonna look like, what the API is. ~Um,~ but it's like in 10 lines of code, you're able to, ~um, like~ do image classification or whatever.

So it's gonna be a really powerful, ~um,~ piece ~to,~ to people building apps with ai.

Paul: So could I like go on to hug and face and find a model? Download it theoretically, ~like~ use the api, upload it to my account, and [00:22:00] then run it.

Greg: it. Yeah, that's exactly, ~um,~ you're gonna need to, we don't have any, ~um,~ training product, but if you've got a pre-trained one, either you've trained it yourself or like you say you've downloaded it from one of these, ~um, sort of~ online libraries, then yeah, you can just take that, upload it to us and we'll run it ~with your,~ with your code.

Paul: ~I,~ I think one thing I love hearing about Constellation is it's definitely like the cloud flare, ~um,~ type of reaction I expect from the AI space taking off instead of ~like,~ let's make ~a~ a product, you're like, let's make ~a,~ a thing that lets people make products, ~uh,~ so we can upload and, ~you know,~ instead of something like gimmicky, you're actually giving a platform for us to make our own gimmicks.

Greg: Yeah, I think, ~um,~ like hopefully not just gimmicks, but ~um,~ Practically every product we build,~ I mean,~ we're building for developers first and foremost, but then we think about how it can, ~um,~ like scale and be used at, ~um,~ like for real in, in proper platforms. ~Um, so, uh,~ workers for platforms ~and, and,~ and pages and, ~um,~ SSL for sas.

All these things are about like, how can we let you build a business on. These things rather [00:23:00] than just like serving directly ~to~ to consumers. We want developers to be able to build apps and then serve their customers.

Paul: One last thing. I think that is definitely, ~uh,~ reminiscent of the CloudFlare reaction, so to speak, of ~like~ the AI stuff coming out is, ~um,~ there is a post that I've seen a blog post about using Lang chain. ~the~ the worker environment in CloudFlare, which is really interesting to me. Cuz ~you know,~ if you develop in workers,~ you know,~ there's only some libraries you can use.

There's only some supported features that, there's a lot there. Almost everything you would need, but you have to pay mine to this and your bundle size and stuff. But here we have Lang chain supported and CloudFlare is really like lending a hand, ~you know,~ people wanting to tool. In a custom way ~with,~ with your platform, are you finding people already jumping on this and using it?

Is it, or is it available right now I should ask.

Greg: ~Uh,~ I can't say off the top of my head, I'm afraid ~if,~ if that particular one's available, but I think, ~uh, if,~ if it isn't, it will be soon. ~Um,~ but certainly there's a massive amount of demand for, ~um,~ being able to use these, ~um,~ [00:24:00] AI tools ~and,~ and, ~uh,~ primitives that we're giving developers. ~Um, it's.~ It's something we're seeing ~like, um,~ being integrated into so many different apps across the board, ~like, um,~ obviously open AI are really ~like~ leading the charge with all the development they're doing. ~Um,~ but we're seeing it in just apps across the board, whether it's just ~like~ a little helper, ~um,~ alongside your, ~um, like~ admin dashboard or whatever that helps you find stuff or whether it's a really rich integration with ~like,~ I don't know, your data or whatever.

~Um,~ but ultimately I think if we give developers ~all these,~ all these pieces to be able to integrate with their platforms, then~ uh,~ we'll be able to see like where all this AI stuff takes us.

Paul: What do you think is the most compelling thing going on with AI right now over at CloudFlare?

Greg: Ooh. ~Um.~

Paul: On, on a personal note like that, ~you know,~ made you ~kind of like,~ Hey, that's actually pretty cool. I like that.

Greg: Yeah, we've, ~um,~ we've been developing a few pieces ourselves, ~so, um,~ We, ~uh,~ we've developed Cursor, which is a, ~um,~ AI helper for ~sort of ~browsing our dots and just ~sort of~ getting started with CloudFlare. ~Um,~ but I also [00:25:00] saw a really cool demo of somebody, um, who'd ~sort of~ explored this, ~um,~ this way to write workers with ai.

And it was this sort of interactive environment where you're able to just ~sort of~ prompt a, ~um,~ Trained model on ~like~ that knew all about workers ~and,~ and how to write workers. ~Um,~ and you could lead it in directions. It was really interesting flow. ~Um,~ and a really, ~if,~ if that sort of way of developing apps takes off, I think we're gonna be in a place where, ~uh,~ you can have an idea and then just ~sort of~ prototype it out real, real quickly and it's gonna really change, ~um,~ the.

Hopefully the accessibility just into development general. More and more people will be able ~to,~ to build apps when they just have an idea. But also, I guess the, ~uh,~ just the way that developers are building apps and just hopefully simplifying much of ~their,~ their workflow.

Paul: Yeah, you mentioned cursor, so is that something that folks can find in their admin dashboard if they log in to CloudFlare?

Greg: Yeah, it's available just in our developer doc, so you can go to, ~um,~ developers.cloudflare.com today, and, ~um,~ Chris should be there. And it's, ~uh,~ just ~sort of~ the chatbot experience for, ~um,~ [00:26:00] finding out about ~the,~ the products we offer and, ~um, uh, ~all the different ways you can use them to, to build apps.

Paul: ~Oh, Greg, we're coming up on time a little bit. We're not right at the finish line just yet,~ but I wanted to leave some room at the end of ~our podcast,~ our conversation today, if you wanted to throw anything out that happened during developer week ~between, you know, the 18th, the 19th, and so on. Even after,~ if you and your colleagues were talking about exciting developments and things that are coming in the future, even if they weren't announced, if you can share anything.

Greg: ~Let me very quickly look, um,~

Paul: ~Yeah, take your time. Just want to do one last sweep to make sure.~

Greg: Yeah.~ So, um,~ there's been a bunch of updates to just ~the,~ the platform in general, a bunch of new primitives. We've had some updates to, ~um,~ D one to the node JS compact that we offer in, ~um,~ CloudFlare workers. ~Um,~ we've even launched a new way to make, ~um,~ Socket connections out of workers. ~Um,~

Paul: Oh, wild.

Greg: to interact with stuff that isn't just available over http.

So I saw some really cool, ~um, uh,~ demos being built just literally on day one of people like speaking to Minecraft servers and stuff and getting information out of them. ~Um,~ so I think all of this is about really building, ~uh,~ powerful primitives, [00:27:00] a unified experience for, ~um,~ bringing them all together when building apps and.

Hopefully just letting us, ~um,~ offer a really smooth on wrap into the developer platform and people can get started ~very,~ very easily. One of the things that we did announce, ~um,~ and I'll finish with this, was the, ~um,~ N P M create CloudFlare. So this is a new C l I that is, ~um,~ integrated with Wrangler. And, ~um,~ that will let you create and deploy a new project, ~um,~ in about a minute.

It's a very impressive thing, ~um,~ that some team adjacent to me, ~um,~ worked on and, ~um,~ You run mpm, create CloudFlare, it will walk you through in a really nice terminal interface. ~Um,~ all that you need to pick from a template. ~Um,~ choose ~the,~ the framework that you want to use. ~Uh,~ so we've got remix, we've got quick, we've got spelt kit.

All the options that you would expect. ~Um,~ we integrate with their CLIs that you're always getting the latest, greatest from those frameworks passes you back and then, ~uh,~ you are set up to be able to deploy with, ~um,~ Wrangler. It's [00:28:00] really, yeah, really great experience. So hopefully more things like that to come.

But yeah, ~very,~ very exciting for the CloudFlare developer platform.

Paul: And Greg, what's gonna be your focus over the next quarter?~ I know you're on the Page's team, so I, I reckon it's gonna be something up that Allie.~

Greg: ~Yeah, so, um,~ like I say, we've got this, ~um,~ team offsite. We're all off to, ~um,~ and we're gonna be focusing mostly on this unification with workers and how we can bring together pages and workers into one singular product. ~Um,~ and hopefully that goes pretty well. ~Um,~ but it's gonna be keeping us very busy for the next couple of months.

Paul: All exciting stuff, Greg.~ I mean,~ CloudFlare is growing to be quite the behemoth in every developer's toolbox right now. ~Like,~ you can find something for everything. ~Uh,~ if people wanted to keep up to date with what's going on at CloudFlare, Through you. Do you post on Twitter or Medium or anything like that?

Greg: Yeah, I am, ~um,~ at Greg Grimble on just about everything. Twitter's probably where I am most active. ~Um,~ and then for CloudFlare itself, ~uh,~ blog.cloudflare.com or our developer documentation.

Paul: Greg, it was a pleasure having you on. Thanks for your time.

Greg: [00:29:00] Thank you very much.