All about Cloudflare’s Developer Week 2025 (AI, vibe coding, agents, and more)

Hello everyone and welcome to This Week in NET, directly from our Cloudflare's London office. In our reception we have one of our walls of entropy. In this case, it's a chaotic pendulum's walls of entropy. In a way, it's contributing to help make the Internet secure. Come with me. And by the window, this amazing view to the Big Ben. Fun fact, Big Ben is actually the nickname for the great bell inside the Elizabeth Tower, which is part of Palace of Westminster, also known as the Houses of Parliament. And yes, I checked this on Wikipedia. To run through some of the week's announcements, we'll have Ricky Robinett and Craig Dennis from our team. And they're just waiting for me here in the Cloudflare TV studio. So this will be an in-person episode. Come. You ever do a pick a card, any card trick? Yeah, the most. Okay, here, take a card. Okay. This guy. And so I'll do this trick. Weird thing about this trick. Yeah. Works on any card except the Queen of Hearts. Magic, Ricky. Hi, everybody. What's up? Ricky's doing magic. This week was magic. I feel like we did a magic trick. I feel like Cloudflare did a magic trick just like that I feel like Cloudflare did a magic trick to us because we have all this cool new stuff. Hello, everyone, and welcome. We're in our London office with Craig and Ricky. Hello. Welcome to our show. Thanks for having us. Yeah, yeah. Excited to be here. So Developer Week, we already did actually an episode last week. For those who want to learn more about, hey, Cloudflare actually has a developer's platform. We did an episode last week where we actually explain and actually Craig explains what the developer platform of Cloudflare is all about. So you can check that episode. This is more about a recap of Developer Week. So for that, how can we start specifically? We are here in London in person, which is kind of cool to be like doing actual in-person things. But before that, what are the key takeaways we can take from our Developer Week full of announcements? There's a lot. I think that's the big feeling of it. It was kind of neat. I'm not sure if all weeks do this, but the way that things are kind of broken up by day, I thought there was like a theme. Monday, AI, agents, Cloudflare being the place to build agents, Workflows GA was there. Yeah, yeah. What else did we have on Monday in that? I can help you there. The man with the hat. Exactly. So let's go by parts. First, we had our Developer Week intro on Sunday, but Monday was the actual piecing together the agent puzzle, MCP, Authentication and Authorization. MCP, yeah. And Durable Objects Feature. But just to give folks a little bit of guidance, first, what is this in terms of MCP, agent puzzle? What is about and what does this enable, really? Sure. Yeah, yeah, yeah. So it's a lot more than just Buzz. I think that there's a whole... We actually had a talk last night that was really great about thinking about how this is actually changing the web and the way that things are working, the ways that things can be consumed. There's agents and this is real. And we have a great platform to come and explore and it got greater. It was already there to have this agent that could kind of live and do tools and do things for you and it got better. Well, first of all, like Ricky said, workflows go in GA. That's something you could trust it now. So you have a long running workflow that you want to make something happen. We have this durable running of a workflow that goes through there. And the agents framework launched before Developer Week, but really landed. I feel like it really landed. We're ready. We're ready to come and play here. I'm really impressed with where it's gotten. Yeah. Well, the thing I love actually on this topic is durable objects. They've been around for a while in the Cloudflare world, right? And we, Craig and I have a lot of conversations with developers of, okay, how should I think about these? What do they do? What should I use them for? And it's one of those strange moments where we had the perfect thing for building agents. We just... Agents didn't even exist as a concept when durable objects was created. And so it's like this moment where it all came together is so beautiful. Yeah. And actually, so Kenton has come and said, when they're coming up with names and they land on durable objects, it was supposed to be called agents originally. Like that would have been like two serendipitous, I think. Yeah, that would have been good. But yeah. So where we're at now, this thing really supports it in a way that you need. And so MCP, model context protocol there. So the way that agents are going to talk to each other in the way that you're... We have tools locally that we're working with AI tools, right? So we have these editors and things like that. And it's allowing you to connect to a set of tools that maybe you didn't even write. You just kind of plug in this set of tools. And now the AI has more abilities that it's able to do. So we have a very nice implementation there. One of the things that's hard about any new protocol is that, how do I do this? How do I get this up and running? We have an excellent starter kit. It just kind of, it goes, it's up and running. And we've thought about the problems because like Ricky just said, this durable object thing, we've been working on this for a long time. So we've got a lot of these problems already solved that you need for this new paradigm. And I'm excited for people to... And, oh my gosh, it's on the free tier. It's on the free tier. That's a huge announcement. One of the things that like, in the DevRel space, I'm like, hey, you should try this thing. And they're like, okay, cool. I can't because it costs... I mean, like our dev platform is $5. But still, I'm still asking, hey, buddy, can I borrow $5 so you can try this thing? But now, just come try it. It's completely free. Really powerful. It's quite interesting. I've done this show for a number of years, speaking about different topics related to the Internet mostly, and AI and computing. But one of the things that sometimes we talk about protocols, IPVC, all sorts of protocols. It's quite interesting to see how old protocols no one cared about. We were talking about durable objects, not a protocol necessarily. But an old thing that was there, no one cared too much about. Suddenly, it's helpful. It has like a second life. So it's quite interesting to see how sometimes tools that were not ready for prime time in terms of people making use of it are now actually doing the difference. Yeah. I think that when people get durable objects, it is this lightning... It's like, I always say that it feels like you shouldn't be able to do what it does. Like, wait, that's not how servers work. So when people touch it, they build incredible things. You saw something at Connect, right? Yeah, yeah, yeah. TLDraw, if you haven't seen this before, it's amazing. Steve, the founder, he showed off what you can do. And it's just mind-blowing. It's a whiteboard application, but the really amazing thing they do with durable objects is collaborative. And because they can connect to people all over the world, it's instantaneous. And the thing he showed, which I had never thought about, is a lot of the ways people do this real-time collaboration is they... Let's say you're dragging an item across your whiteboard. They get the start state, and then they see it's moving, and they don't communicate that to the other users until it stopped moving. And then they send that update to everyone, right? That's how people typically do this. Because they use durable objects, they can send everything as it's happening. And he was talking about just the fidelity of communication you can do when you're not dropping information on the floor because your infrastructure can't support it. It's so wild. And quick, right? Because making it quick is quite important. Like, latency there, quite important as well. I'm curious on... Monday was actually a day with different announcements. You were mentioning workflows being GA, generally available right now, so it was already there. But now it's really open. Yeah. And, sorry, real quick. One of these things with agents, right? You don't want, like, okay, yeah, go ahead, run my life, agent. Go do everything. You want it to stop and ask you, like, hey, I'm about ready to do this thing. Is this cool? You want that step. So we added this wait for event that was part of this GA that went. So you can say, wait for the human to say, yes. And then it continues on. So it's super powerful. And again, one of those things that everybody needs to do it. And it's like, wait, how do I do this? I mean, if you think about it, how do I write a server application that's going to wait for so much time? That's going to cost a lot of money. But here it doesn't because we let it sleep. We let it go to hibernate and come back to life. Yeah. Well, and I love this because it's like, wait, it could be one second. It could be days. Yeah. And so as a developer, like, how do you solve that problem before? It's just so, so tricky. Yeah. We didn't mention still AutoRAG. That was like introduction. For those who don't know, what is AutoRAG? Fully Management Retrieval Augmented Generation. Yeah. RAG. What a cool name. Yeah. I'll take this one. But for Craig, just fill in any gaps. So if you haven't worked with RAG before, the idea is, hey, you have a large language model. It doesn't have the full context about your business or yourself to answer all your questions. So this concept was developed of Retrieval Augmented Generation. RAG. And so what happens is you take your content, let's say it's your emails. You load that up into what are called embeddings, which are ways to store that text using math. You put that in a vector database. Then when you are talking to your large language model, you retrieve what's relevant to the conversation. And all of this, I don't know, it took me a long time to wrap my head around it. And there's a lot of moving pieces to set that up. So AutoRAG, as the name suggests, takes the heavy lifting out of that and lets us handle that for you. So you can just give your LLM or your AI agent that context straight away. So you give us your documents. We do all the embeddings, handle the vector database, and then you just have an API that gives the conversation back and forth with the correct context. It's really kind of mind blowing. I remember when we started talking about this, I was like, cool, I don't know if it's going to work, right? And then when we started seeing it, everyone was just like, this is it. This is what we wanted. Very interesting. Yesterday, actually, Rita Kozlov did a demo at Connect here in London with AutoRAG. Also, our new, what is it called? The, not agent, but the assistant on the dashboard, the name. Cloudy. Cloudy, that's it. Cloudy. Cloudy was also there in the demo as well. It was quite interesting to see it work, to see it in the dashboard just working and with Cloudy helping, actually. So if you don't know too much about all of the topics, having something there, in this case, a chatbot, helping is quite interesting. And also, it doesn't do everything for you. You were mentioning that. It asks you, this is the setup. Yeah. And do you want to do this, right? Exactly. So that's quite interesting to see. And so if you're at a company, you've got some, maybe you have some white papers and you have stuff that you want your tools to know about, maybe it's a PDF. Maybe it's something that typically, as a developer, I'd be like, okay, well, I could probably find a library to figure out how to do this. I could go, but now I can just go, bloop, and it's auto, right? And then it's part of the search that you start using. Well, and why is it, for me as a developer, every time I have to parse a PDF, I think, oh, this is 15 minutes. And then three days later, I'm still trying to write the code to parse out the PDF. I don't know what it is. It feels like an easy problem, but it never is. It never is, yeah. I've been there. I'm not a developer, but I've been sometimes trying out things and I've been there. Well, on Monday as well, Cloudflare acquired Outerbase to expand database and agent developer experience capabilities. What can we say about this? These guys are awesome. On social media, sometimes you're like, wow, they really like Cloudflare. Wow, they're really building this awesome tool. I wish I had that. I'm going to use this other tool to help me understand this application I'm building. I mean, I have. I have gone and tried to use their tools to look at my database the way that they're presenting it. It's gorgeous. They do gorgeous design and developer experience. They really think about, hey, you might want to know. For instance, one of the things that they've built is durable objects are sometimes hard to know what's going on inside of there. They built this really gorgeous way where I can go and query and see what's happening. How long has it been up? What's data inside of here? How many connections do I have? It's just this really gorgeous DX. They're great guys. They ship. They're wild. In fact, they ship something. I mean, we'll get to it here, but they shipped something this week. We acquired them Monday and they shipped something this week. So amazing. Well, it's one of those places that we all in chat have the companies that every time they ship something, we're like, did you see this? And I don't know if I can name a company that we do that more with, where it's like, did you see this? This is amazing. And so I remember when we started talking about this being like, please, yes, we want them part of the family. Yeah. And like, as somebody who doesn't know, I literally air punched. I jumped in the air. Yeah. So really excited. I got to talk to him yesterday too on a stream. They're great guys. Great guys. Let's go to Tuesday. Tuesday was a different day about MySQL apps using the call flow workers and Hyperdrive. Full stack. Full stack. This is the full stack day. What can we highlight here? Also, Pools Across the Sea, actually a great name for a vlog. Pools Across the Sea, how Hyperdrive speeds up access to databases and why we're making it free. Yeah. I forgot about that one going for you as well. Hyperdrive is so cool. What is Hyperdrive? Yeah, yeah, yeah. So I'll take this one. So when you go in a typical application, you have to connect to your database. You have to go, hey, how's it going? These are my credentials. This is what I would like to do. And it takes time, right? You need to go do that every single time. So in a serverless environment, you have to like, okay, you don't hold on to anything. You can't hold on to that thing, typically, in a serverless environment. So what we've done is we've made a thing that holds on to that connection. So you can do a connection pool to a whole bunch of databases. We added MySQL. And what I like about Hyperdrive specifically is the way that we listen to customers. Customers are like, I need MySQL. I'm like, okay, MySQL. It really is great. During this day, I did a live stream. John Resig from UpjQuery fame was on. Huge fan. It was a moment of my life. It was wonderful. And he had a specific ask for a database on chat. And Thomas was like, oh, we can think about that. So it connects to these different databases. Your code doesn't change much. You just change the connection string. But the amount of how much you can save. Of course, we have other stuff on top of that. Now, if you're running through this connection that we have and we're keeping open for you, you can cache. You can cache queries as you need to. And it's smart about caching queries. And it also has smart placement. I think something happened with smart placement, I think, this week, too, where it can put it as close to the database as it needs to. It's incredible. If you look at the graphs of what people were doing versus what happens, it's one of those really nice, if you try to show how long it used to take. It's like, really big. One of the things I love about this, but this day in general, for me, I always want developers to be able to use the tools they know and love. And one of the things that's great about working at Cloudflare is I think we try to make our platform as useful for as many developers as possible. And so, the HyperDrive stuff is amazing. MySQL is great. We are huge fans of PlanetScale. And so, that was one of the first things I think we were all hearing, like, hey, we want to use workers plus PlanetScale. And we're like, now this is the way to do it. Also, the VEET stuff, which we haven't even touched on yet, is another one of those things of, hey, we hear from developers, this is our tool chain. We're using VEET. We're using MySQL. And making that all easier always brings me joy because we don't have to tell developers, you've got to use this specialized Cloudflare thing to be able to do it. Yeah, yeah, absolutely. And I think, too, the level of like, you might have been able to figure out how to do that yourself, but that's not what you're supposed to be building, right? There's time as well. Yeah, it's time. It's a challenge. But really, you probably should just be focusing on the thing your company is asking you to build or the startup you're trying to build. You shouldn't be needing to focus on that stuff. And I feel like the VEET plugin, it really brings people there. Because when you go to build software, especially front-end applications, it's gnarly. It is really challenging. And there's been flame wars almost across the Internet about, you should use this system. And people are finally at a place. I feel like there's this really nice Switzerland piece place of VEET. And VEET means fast in French. And it builds really fast. And you don't need to worry too much about anything. And you just do this plugin. And now you can access all of the bindings that are available. So you can connect to everything now. So it's really powerful. The... Everybody internally, right? So people who have to build in... We dog food a lot, right? So people internally are so excited about this because it's been a problem in the past to figure out, how do we build all this stuff? So then we can kind of standardize and feel good about that. Well, and the local dev production thing with VEET is so magical, really. And I think that's one thing that we keep getting lost on because I don't think we want to admit how exciting it's going to be that we start seeing that, I guess, paralleled in a way we haven't before. Yeah. And so VEET comes with a lot of cool stuff too, with the hot module reloading. If you're working on something and you can imagine, you're trying to like, okay, so I need to use this and now I need to deploy it and did it work, right? So you need to... Okay, so I'm going to do this and then I'm going to do this and then I'm going to do this. Okay, there. And now it didn't work. Okay. So now I'm going to come here and I'm going to do this and then I'm going to do this. And that's not VEET, right? That's not fast. It's not fast. So now it is. Now you do it locally and you can see what it feels like, exactly what it's going to feel like. That's the beauty of this. It is exactly the same because we're passing it the way that we're passing things through. Because we worked with the VEET team, I was on a screen with the VEET team and they were loving the fact that we helped them move this environments thing forward so everybody can play in this space, right? Because I really, I love to help build a better Internet and VEET is helping build a better Internet and we're helping VEET help build a better Internet. So I just feel like this is the help build a better Internet day. Yeah. You can also deploy now static sites and full stack application on call for workers. That was another announcement specifically. The framework support for React, Router, V7, Astro, Vue and more are generally available today as it is the VEET public plugin that we mentioned. What can we say more about this day? I'll just put one more plug in this day. The deploy to workers button is something that we had kind of broke, didn't use it as much and now it's back and it's the best it's been. But also, if you had the deploy to workers button, the previous version, it just started working again, which is kind of great. I mean, maybe shouldn't have stopped working, but I love that instead of just replacing it and telling developers, Oh, go back and update all your code with this new thing. We're able to say, Hey, we have a new thing. It works really good. Also, if you were using the old thing, that's just works again. It's really, really powerful. And so you're going to see, especially stuff Craig and I and the team do having that deploy to workers button everywhere. So you can just get started without having to do all the configuration. You can get straight to coding. Yeah. Yeah. And that, that, I mean, sometimes you want to build, we want to, we want to show you what a, what a production grade application feels like, but sometimes it's like, all right, in order to use this, I want you to run these 17 commands to go is you're going to set up your database and then you're going to set up your vector database. And then you're going to set up your auto rag. You're going to, you know, but now, uh, they, they've made it so that it provisions it. So, so you can set your, your repo up and everybody at home can set their repo up this way too. Like if you're, if you are trying to show somebody or trying to show how your tool interacts with this thing, you can set up your, your configuration and a sort of half a half, like a part baked, I guess is probably the right, the right term for it. Uh, it's, it's there and you, you, you deploy to, uh, workers and it goes into provisions, everything for you if it needs to. Uh, and I, I happen to wrap my brain around how far we can take things now because oftentimes you want people to, you want to get started on what you're trying to show what they're doing. So oftentimes we kind of back off a little bit, but now we can actually put the whole. True. To be honest. And sometimes I do a bit of that and having like, okay, sometimes I just save it. Oh, okay. Here's a set of steps. I'll save it. I'll do it. Uh, and another, another time and having like a button that actually does the step, it's like an agent type of thing. Yeah. So you go to the dashboard and just is there already the status. So it's quite amazing to see, to be honest. Yeah. Yeah. It connects to your get hub too. So like it'll, it'll take it. It'll even put it on your get hub for you. So you'll have this like level of where it starts at and then also deploys from there. So it's, it's really, it's, and the topic for this day is like call flow is zero friction and it's zero friction there as well. Uh, Wednesday, uh, platform for enterprise day. Um, went over said that's nation for any workload. That means enterprise is ready with the security of civility and performance capabilities. Yeah. What can we say about Wednesday? Uh, real time kit, real time kit. Uh, real time and real time kit. Yeah. Yeah. Uh, so, uh, for anyone who's built real time voice and video applications, Craig and I have spent basically a decade plus in this space. Uh, so we're especially excited here. Uh, it's really, really hard. And it's one of those things that looks easy. I think as developers, we've all seen the, it's just 15 lines of web RTC and suddenly you have a peer to peer call and, and then you actually want to deploy it and you start realizing like, Oh, it's not just the peer to peer call. And there's people all over the world. And then you start being at the constraint of the person with the worst connection and, and all these problems come together. And so, uh, we launched calls, uh, I guess maybe a year or so ago and have been in this real time space learning so much. And so consolidating that into a toolkit that makes it really easy for developers to deploy. We acquired a company called Dyte that, uh, happened before developer week, but this was kind of their, their coming out of what you can do with it. And if you watch the stream, uh, whenever we talked about this, I guess Wednesday, uh, one of the things I loved is in the docs for real time kit, they have an embedded code snippet that runs real time kit and you can experience it live. So you don't even have to deploy it yourself. You can go to the docs, you can read about it and then just start using it live to see what it looks like. Oh, that's super cool. Yeah. Yeah. And, and I, you know, this is coming, right? Like people talking to your computer is coming. And, and the more, the more you're able to do that, if we can again, reduce friction and like make that easier, I think we're going to see cooler apps come out of this. And of course, you know, like this is also running across our network, right? Our, our global network. With benefits in latency, those types of things, which is important as well. And it's truly important in that. And it's truly like this interactive nature that is going on of you talking with computers really, uh, that is really becoming something here and working because if it, to be honest, if it does not work well, if it's lagging, if it has problems, the experience will be much worse and people will use it less. We've seen it in the Internet as in general, uh, and we can see it in applications. We can see it everywhere. Actually, this was also the day about introducing Koffler secret stores, better secure secrets, simplify your workflow. What is secrets? I don't, I have seen how excited people are, but I haven't dug in. Did you, I, I could touch on a little with this one. Uh, people, it's funny the things people get excited about, right? Uh, so historically when you store a secret with your worker, right? Like that is stored with that worker and you know, it's kind of isolated into one spot, which sounds great. And it is great for a lot of scenarios, but secret store is this higher level abstraction where, Hey, there's some stuff you need stored across multiple projects beyond the developer platform. Right. And so being able to store these secrets in a way that's accessible, uh, multiple places is really powerful. Uh, and so I think folks have been, uh, wanting that level to abstract it up a layer. So that is my, like, again, uh, like Craig, it's like, there's so much, uh, every week, this is one that I'm like, you know, just an inch deep on, but that seems like where people are getting really, really excited about it. There's also, I think that makes across agents. Like, so, so these agents are going to need to know it's like a pocket of like, this is a very secret place where we're going to, yeah, that makes sense. Also call for snippets are, is now generally available, um, enabling fast cost-free JavaScript based ATP traffic modifications across all paid plans. Yeah. Uh, I can talk about snippets too. So snippets, uh, a little history on workers for people who haven't heard the story, uh, you know, the way that we started workers was, Hey, we had this incredible network. Uh, people were using the CDN, they'd come to us and they'd say, Hey, we want this customization. Right. Uh, and it didn't really make sense to build, you know, a button in the dash for every single thing people want to do. You suddenly have this like giant world. Uh, and, and what we said is, Hey, we can give you this layer, the worker, and you can start just coding your set stuff yourself with JavaScript. Right. Uh, really, really amazing. It's grown into this entire developer platform that we see today. Snippets is a little bit of a return to that original observation to say, Hey, there are actually a set of things that people need to customize when they're using Cloudflare. And snippets let you do that really quickly. And maybe you don't know how to build a worker, right? Or you're not a developer. And so it's kind of that in between to, to cover those use cases and give that customization still that you wouldn't have if it was just a toggle on the dash you were turning on. That's nice. Yeah. I bet. I bet that's really nice for, that's a nice introduction to, to workers. Right. You're like, Oh, that's cool. What else can this? Oh my gosh. Yeah. Peel back. There's a lot of talk about observability, introducing workers, observability, logs, metrics, and queries all in one place. Yes. Uh, so, so good. This is another one of those sleeper hits of the week. Uh, we'll talk about feedback at the end, but, uh, I will say we, we were at the celebration in London with a couple hundred developers last night. And the amount of people who pulled me aside and said, my favorite launch this week is the observability, uh, stuff, uh, blew me away. Uh, so, uh, incredible dashboard for looking at your logs. We, we launched, I want to say workers logs last year, sometime like leveling that up, but also we acquired baseline last year, uh, continuing to integrate all that great observability expertise. And so now if you want to look at your workers, you have this incredible query builder that you can dig in and really get to the heart of what's happening on a worker. Yeah. And when else has this happened? Right. And so like, Oh, there's an error here. Oh no, I didn't know about that error. How many other times has this happened this specific and to just be able to search and find that. And that's what, I think that's what people were talking about last night. I was like, Oh my gosh, I've, I've wanted that forever. Cause you normally have to just go through, right. But now you can query into your, your space. Yeah. If I, if I were to plug, uh, Rohan jumped on our stream, uh, on Wednesday to talk about this and he does this incredible demo and it's just, yeah, you can go in and you see one thing and suddenly you're spelunking through all your logs to find the patterns. And, uh, at least for me, that's the way I use observability tools. And I think it's going to unlock so much, uh, really, uh, debugging and fixes, uh, that people may not even realize they need on their worker. Yeah. Yeah. In many areas, actually, not only developers observability is important, but in this area, it devolves like not only bugs, but it saves you time. It saves you sometimes even it's not like only bugs and time is actually big problems. Like something that doesn't stop working and you don't know why. And you're just wasting time with that. Uh, on Wednesday, we also, we did this, we do this, uh, every now and then, uh, network performance update, uh, regarding our speed in terms of network. So something general, but it's always present because, uh, workers uses the Cloudflare network, uh, and benefits from it. Uh, why not go to Thursday already? Uh, yesterday. Um, and the first announcement I have here is making super slurper five times faster with workers. Good job. You did it. Yeah. It's not easy, not easy, especially in English. Not your first time. Yeah, definitely. Yeah. So super slurper is how you move data, uh, between, uh, places. Right. So, so if you need to meet between different cloud storage places, so, uh, be it being five times faster, of course makes that process a lot better. And, uh, yeah, it's, uh, super slurper is a great tool because you might want, I'm going to, uh, I know, I know it's coming up in the rest of the day. So I'm going to bring this forward here. There's a, there's this thing that happens. It's kind of strange, uh, in, in cloud providers, especially with data, you can store your data in a place, right? You can go, you're like, I'm going to put it, I'm going to put it there and I pay for that. But sometimes when you go to get it, you go to pick that data back up and use it somehow. You also have to pay for that. And that's kind of weird. It's, uh, Alex on the streams said this really cool thing where he's like, it's kind of like an ATM fee where it's like, I'm going to get, it's my data. I'm going to get up. But, but, but like based on how much you pull out, you kind of have to pay. Yeah. And that's called egress. Um, and our, when to do that, to use, and so we are using a lot more data than we have, like traditionally, like, uh, because of, uh, AI training data and all sorts of things along those lines. Right. So there's a lot more data that's out there. Uh, we're using data to make better decisions and big train models and things like that. So we, you often want to move to a place that doesn't have that fee. And that is us, right? So super slurper helps you do that. It helps you like pull from there and it being faster, five times faster, it was fast but like five times faster than what fast was. It's really fast. Um, so, uh, and then it gets you to move it to the right, the right place. And then I think it also, I think super slurper is, I'm, I'm a little over my skis here. I think super slurper lets you choose to use it only when you want it. Like I, I, I'm going to, I don't want to move all of my data at once. I'm going to go get my data when I need it, if it's someplace else, but if it's here, use it here. Otherwise go get move it. And then at that end, so then you, you, cause you, you don't want to pay for all of the egress to move it to someplace else. Right. So you can kind of conditionally move the, the data across there. So that's, that's super slurper, which is a slurpee icon, which is, I'm so glad that that exists in the Cloudflare emoji. I feel like no one says super slurper without smiling. That is, that is my favorite thing about it. It's a great tool, but it's also just a good way to, you know, get that little curl. Cloudy as well. It's positive. You can see like a positive thing there. R2 data catalog, manage a patch iceberg. I'm going to stay over my skis, but I learned this stuff yesterday on stream live. So I can talk a little bit about this. Do you know this stuff? I'll give just like a high level and you all get over my skis and then you can go. Which is, you know, so much of what we build is informed by our customers and what they're doing. And our customers are amazing because even if our products don't do something perfectly, they'll find a way to make it work. Right. And so I think basically from the moment we got R2 in hands of people, they're like, I want to use it as a data Lake, you know, things like Snowflake, I want to connect it to. And so they just started doing it. I remember I've been here a couple of years. I remember like one of the first conversations I had with one of our investors was asking about this or like, I read a blog post and someone like wrote about how they hooked up Snowflake and R2. And is that, is that a thing? And I'm like, yes, that is a thing. And so this really, to me is about taking that use case that we've already seen and that customers are pulling together in various ways and saying, Hey, we're going to make that easy for you. And we're not going to make you hack together that connection. We're just going to give you this out of the box basically. And you don't have to think about it anymore in the same way. Yeah. And there's nice, there's really nice standards in this, this space. And so iceberg is one of these standards that is supported so you can connect to these other tools to can connect to it. And again, this is the big data that we're talking about. There's a lot of data here and that you don't, we don't charge for when we pull that data. Well, zero regress freeze is one of the call flares mottos I've heard over the years. And this is also all about that global data experience, make it less expensive. So with this, what's really nice about it is it puts this layer of being able to access and find that data. And we saw some really cool demos. I recommend everybody watch the stream that we did yesterday for this to plug that. Cause they did a really excellent job of walking through it. But you drop tons of data and then run a query over the top of it and find the data that you're looking for to bring that down. And you're about ready to walk into what the, so that's reading the data, but how do you get the data there? Well, we want, so, so that's beta. So, so our, our two data catalog, you, you, you go and you flip, flip the switch on. Now you have this, now you have a data lake, right? So but you need to write to it. And so that's actually hard. That's, that's a hard thing to do as well. So also announced in beta people have been waiting for this for as long time as pipelines. So you often want to write you want to write data and store data. And I, I got, I talked to the guys, they're both, both guys. So I talked to these guys yesterday about how, how do, why would, when do I use this? Right. So a lot of times like you want to like research like, Oh, this person clicked this thing on my website or, or I built this feature and I want to make sure that people are using this feature. And so you want to push this analytics through. And naively, you might put that in your like production database, right? So like the users are paying for you tracking the users and that gets, it gets gross. And so we didn't really have a way for you to do this exactly into a data lake before. So we, so now we have this pipelines product and you can write to it. And I watched them set this up. It was like instantaneously, there's a binding for it. So you just write down this you just write down this pipe and then also provides an HTTP interface. So you can actually just do that. So you can not even have a worker's project. It could be your project at all. And now you have this pipeline again, have this global network. And again, it's going to go into the global network and it's going to be able to be accessed there. And we do all the caching and the batching and all the, all the hard, all the stuff that makes this hard and you need it to be fast, right? You don't want it to stop. You don't want to slow anybody to feel that you're, you're slowing down. So it was really cool tool. I was, I was blown away by the demos and then like also like, oh boy, if we're going to start freaking out, we're talking about this yesterday. I need to learn some big data stuff. We're going to start, we're going to start doing this. I need to become an expert in this and understand how to, because, and, and to the, to that point, I am not an expert in that because I haven't had access to the tools to like, think about what those, what those are there. So it's nice to, it'll be nice to, to, to do that. And, and to bring data scientists into this world, right? Like, I feel like that's a clever group of people and I would love to have those builders here to build in, build in solutions. I work with a few of those actually with radar, actually. The, are you going to start using it in Iceberg? Yeah, we need pipelines. And we've acquired also a Royal enabling us to bring new SQL based stateful transformations and pipelines and R2 specifically. So another announcement there. Do you do more on that? I mean, I think these are ones that we're like learning more about this week. But what I would say is for me, it's easy to have a naive understanding of all this data and you're, oh, you're just shuffling around ones and zeros everywhere. And it's kind of, yeah, I copy and pasted a bunch of photos I had from my laptop to an external hard drive, but it's a bit more complicated than that. And as you're moving the data around, you do want to make changes to it. You want to manipulate it. There may be data you store for a certain amount of time, but then when you ingest it into your data lake, you don't need all of that or you need it summarized or organized in a different way. And so again, over my skis on this one, but the way I think about this is we're going to get really, really good of not just a pipe that ingest what you're doing, but lets you have that control of what exactly do you want to end up in your data lake. Also have a technical deep dive about D1, Cloudflare's managed SQL database announces read replication better. Yeah. Talk about that. This one ripped. So that was just yesterday. So this one ripped through a crazy amount of traffic on how excited people are for this, right? Because you want to get your database as close as possible to where the people are because you want the latency, right? We've done a really good job of getting you there, of getting the page is going to be served there. But if you can also get the database really close to where the user is, that's where this read replica happens. There's an awesome and incredible demo built by the outer base guys. This is a little callback to when I said they built something and released it. Replicas .pages.dev will show you what it feels like latency wise in different places, because your users are hopefully all over the world. When your app takes off, hopefully your users are all over the world. And this shows you what it feels like by using a read replica. All sorts of really clever stuff too. This is a really hard problem. The replication of data is a really hard problem. So if I go and I push something forward and then I go query it right away, I want to see what I put there. So there's this hack that sometimes you do where you hold some stuff back and you're like, okay, here's the query that came back and the stuff, right? And you have to build that yourself. But this tool has it. There's this thing called bookmarking. It was like a sleeper hit. A little sleeper hit there. And you can kind of include like, hey, they just did this. So when they run this query again, include that data no matter where, if the data is there or not, go get that data back. So it's really neat. Let's move for Friday. Today is actually a startup day, startup spotlight, building AI agents and accelerating innovation with cohort number five. We actually have our supporting startups building stuff. Yeah. I got to be part of that last one. It's such a cool program. So yeah. We also have a update on startup program in terms of their journey. There's a global virtual private cloud to build secure cross cloud apps on workers. What can we say here for this day? It's right now. It's happening. It's right now and we're in London. So this is, Ricky and I are not used to this. We're in the future. Yeah. It's being published in 15 minutes. Yeah, exactly. So I haven't read this one yet. Yeah. I would say, so first startup program, if you are a startup, you're watching this, please, please reach out. We'd love to talk to you whether that's about being in Launchpad or we have credits for startups, even if they're not in Launchpad. I think one of our favorite things is seeing what the most innovative companies in the world are doing with Cloudflare and that's often in startups. So just huge plug for that on the VPC stuff. It seems cool in connecting your worker to other places where you are running things is like my basic understanding of it. So one of the nice things about Cloudflare is I don't think we often are like, hey, you have all this code somewhere else, move it to Cloudflare today. That's just not how I work as a developer. And I think not what we ask developers to do. But one of the tricky challenges has been, hey, I have some workers now and maybe have stuff running somewhere else. What do I do? Do I have an HTTP API build just for connection there? And so trying to make that interaction easier is the high level of my understanding of this blog post. There's a few others. Jump in if you want to say something about it. Workers.ai gets speed boost, batch workload support, more low -res new models, and a refresh dashboard. That team's been cooking. There's all sorts of new models coming, which we stored them up for this week. We've been waiting to have them. So really cool models are coming right now. There's a new Mistral model that's coming out. We were planning on doing a new Llama launch, and it ended up coming up a little early. On Saturday, right? Yeah, it depends where your calendar starts. Yeah, that's right. Maybe it's developer week when we start that. But we launched, we actually had a Sunday blog post about it. Yeah, we did. So the team has really been cooking. Yeah, so there's this batch API where sometimes you want to run a bunch of API requests, a bunch of AI requests, and you don't want to wait. You don't want to sit and run and loop, get the thing back, store the thing, do the thing. You want to get all the data back and then go do something with it, right? So I think you've probably seen applications where there's a spreadsheet, and it's like, I want you to go research all of these companies, and I want you to go. So this is a nice way to use the AI tools in a batch, and then it all comes back for you, which you could do. You could do that yourself, but you don't need to. So we saw that as a common place that you needed to queue up all of your requests and then get all the results back. And I think we'll see a lot of powerful use cases with it. People do that a lot with, we were talking earlier about RAG, you need to do a lot of embeddings all at once because you have a lot of documents, and you don't really want to sit there and do the loop. You can. You can do the loop, but you probably want to do it faster, and especially if it's something that's happening every day. So the batching of these requests all at once is something that gives you peace of mind that it's not going to fail halfway through. Oh, now I need to go run them all again. It just kind of runs them all. And then you get all the responses back, and then you can kind of go associate them. Makes sense. Full week of announcements. Let's go to the feedback perspectives. We had a lot on social media, to be honest. I've seen previous developer weeks. This developer week, we've seen relevant folks from actually building stuff and having their own companies really being wowed into some of the announcements there. What are the most surprising, you would say? Yeah. I love this because it's this validation that we've listened, right? I wish you guys did this. Oh, you guys do that? Yeah. That cycle is really nice to see, having seen that. This is my second developer week of watching this happen. And you see that. You see developer week happen. There's always, and I'm sure there is this time too, people like, this is so awesome. Oh, I wish it did that. And I know that the next developer week, we're taking notes. Yeah, exactly. We're always taking notes. And I really like that flow of this and it's incredibly well-received. Yeah. I mean, just kudos to all the people building these products and these features. Craig and I have the privilege of getting to kind of take them and show them off, but there are so many people building the stuff and thinking about what to build. Plus one to Craig's point, and one of my favorite things, and I don't know if it's surprising, but it's encouraging, is when we show something like observability and people are like, okay, this is great. Here's 10 more things I wanted to do. And I think you could be discouraged by that. But what I love about our culture at Cloudflare is it's actually an indication that we're on the right path. Because if you demo it, people are like, oh, cool. Thanks. Thumbs up. It means you haven't sparked their imagination or shown them what's possible. And so I love when we show something. And I think auto-rag is another one of those things where it sparked people's imagination. And then they're like, hey, could it do this? Could it do this? I wish it supported this file type. And so those are always... I shouldn't be surprised, but maybe I'm surprised by how passionate and engaged our customers and community are during developer week and all the time. And the trust that I think our product and engineering teams have built that people know if they say that stuff, they're not going to get dunked on or it's not going to be drama. It's going to be adding it to the backlog of things we want to build. Yeah. And I think there's a really nice appreciation. You must, if you're working in product here, you must feel how much people are like, that was so much stuff. How did you do that? Yeah. So I hope that product feels that, because if I were to echo what the community is saying, it's like Cloudflare ships and listens and ships. And I think that's the main. And I hope that y'all have a nice long weekend, all you product people. Absolutely. And for example, even yesterday, I was messaging Gabrielle from our team about Autorag and the demo. And you can see the clearly need for feedback, how it was received and how can we improve. It's quite interesting to see. And I get to work with some of those folks in our Lisbon office specifically. So quite interesting to see. I want to, before we go, just to get a perspective on some of the topics, we already spoke a lot about agents. In terms of agents, vibe coding, this amazing little thing that has a term to it. What are your biggest expectations for those really? Not to get too philosophical. I'm like trying to figure out if I'm going to get existential on us here. Do it. Bring us there. Yeah. We're doing it. I think that humans were made to create. I think that when I see the people around me filled with joy the most, it's when they're creating something, whether that's music or that's software or that's art or, you know, humor with jokes, like that is all a form of creation. And when I think about what's happening with AI, I get the people that are cynical and it's really easy to be cynical. But what I see is the ability to create, especially with code being put in more people's hands. And like, there's this meme as a developer of your family and friends saying, hey, I have an idea. Like, can you build this for me? And we're entering this space where I actually feel like I can say to a lot of them, you, you can build this for you. And I think that's beautiful. And I think that's going to change the dynamic of how society works when more people can take things that are in their brain and bring them to a reality. And so when I think about where all it's heading, I think it's just this explosion of creation in a really beautiful way. And I think it's such a privilege that we at Cloudflare get to be a part of that with the developers and future developers around the world. Yeah. Yeah. Yeah. Wow. That was, that was great. That's a clip. You can put like a, you know, fancy, like, I don't know, philosophers. Yeah. I saw somebody recently said that they were trying to figure out how to build good slides for a talk that they were doing. And instead they vibe coded a slide engine and had it styled and had the, the, the, what he was trying to say with his voice. And it's just like, okay, like maybe that's where we're at. Like, you know, like maybe we've gotten to, maybe we've gotten to that place that we've all dreamed of this. Right. And I think, I think there's this, there's this like, we're a little hesitant because we've seen, we've seen the movies we've dreamed. People have been dreaming and then it becomes real, right? Like this is what the future could look like. We see the movies, we implement the movies. Right. And I feel like we're at this, we're at a stage where this, the future is right here. Right. And I find it also interesting, the fact that sometimes more than the movies, sometimes we didn't imagine how it could be right now. And it seems that it's getting there. Sometimes people say like the last 5% are the most difficult in innovation, but it seems really that we're getting there. And for someone less technical than you guys, what I've seen is the amount of excitement about vibe coding, seeing people build, even I've seen experienced people build with no making their magic, just using like AI tools. But I've now started to see people less, less experienced or developers at all, just start doing it. And you can see like the magic type of perspective there. And it's, there's a really excitement in the community. You can see it. And I think we get to see it in coders, but I think the rest of the world outside of this is going to have this experience. We're just getting it first. And I'm excited for that. And I hope that we're able to, we're able to help build that, to let people build those things that enable people. Right. It reminded me a bit on when ChachiPT was launched. I actually did a This Week in Ed episode with John Graham-Cumming. And we actually did an episode where it was just it was so big, not in the whole world. In those first two weeks, it was more developers, people in technology, but you can see like it's fine. So big. We did actually an episode about that. And some folks actually told me, you're speaking about that thing that will pass, that will pass. That's like a fashion of sorts. And I said, I don't think so, but good luck with that. And now actually, I sometimes reference that episode, just we did a bit of actually coding with those early models. So it's quite interesting to see evolution. Yeah. Only like two years and something and changes. Yeah. Quite interesting to see. Yeah. This was great. Any final notes that we didn't say and should? Well, I think if we missed anything, there are live streams that we did all that's available. I think there's a wrap up on Monday. Yeah. I mean, come on Discord, come on the community, discord .Cloudflare.com, community.Cloudflare.com on X, you can get us. And the one thing we didn't touch on is it's not over. I know at least a couple things are coming next week. And there's been this joke of maybe it's developer month or developer year. And so we're going to keep shipping. And so keep giving us the feedback. And we hope we keep getting to show off rad stuff that you want. Exactly. And for those who want to read all the blogs, blog .Cloudflare.com. But also, we have the developer week hub, Cloudflare.com slash developer dash week. So it's all there. The segments that were recorded, video segments with demos, the blog posts, the press releases, all there. So you can take a look. And that's a wrap. Thank you so much. Thanks for having us. It was awesome. It's done. Awesome. AI, agents, Cloudflare being the place to build agents. Workflows GA was there? Yeah.