The Changelog: Software Development, Open Source - When vibe coding goes viral (Interview)

Episode Date: June 27, 2025

Chris Anderson joins the show. You may recognize Chris from the early days of CouchDB and Couchbase. Back when the world was just waking up to NoSQL, Chris was at the center of it all, shaping how developers think about data distribution and offline-first architecture. These days, Chris is working on Vibes.diy and Fireproof — tools that make one-shot app generation not only possible, but shareable within minutes. We talk about the origins of CouchDB, the fork that led to Membase and Couchbase, and how that long journey led to this new paradigm: Vibe Coding.

What's up friends welcome back, this is the changelog featuring the hackers, the leaders, and those who are vibe coding. Today Jared and I are joined by Chris Anderson, a naming man recognized from the early days of CouchDB and Couchbase. So back when the world was just waking up to new sequel, Chris was at the center of it all, helping to shape how developers Chris was at the center of it all, helping to shape how developers think about distributed data, offline architectures,

and now Chris is working on vibes.diy and fireproof, where one-shot app generation is a reality and shareable with friends in minutes. A massive thank you to our friends and our partners over at fly.io. Launch your app in five minutes our partners over at fly.io. Launch your app in 5 minutes. Learn more at fly.io. Okay, let's vibe. Well friends, Retool Agents is here. Yes, Retool has launched Retool Agents.

We all know LLMs, they're smart. They can chat, they can reason, they can help us code, they can even write the code for us. But here's the thing, LLMs, they can talk, but so far, they can't act. To actually execute real work in your business, they need tools. And that's exactly what Retool agents delivers instead of building just one more chat bot out there Retool rethought this they give LLM's powerful specific and customized tools to automate

the repetitive tasks that we're all doing imagine this you have to go into Stripe you have to hunt down a chargeback You gather the evidence from your Postgres database. You package it all up and you give it to your accountant. Now imagine an agent doing the same work, the same task in real time and finding 50 chargebacks in those same five minutes. This is not science fiction. This is real. This is now. That's retail agents working with pre-built integrations

in your systems and workflows. Whether you need to build an agent to handle daily project management by listening to standups and updating JIRA or one that researches sales prospects and generates personalized pitch decks, or even an executive assistant that coordinates calendars across time zones.

Retool Agents does all this. Here's what blows my mind. Retool customers have already automated over 100 million hours using AI. That's like having a 5,000 person company working for an entire decade. And they're just getting started. Retool Agents are available now. If you're ready to move beyond chat bots and start automating real work. Check out Retool Agents today. Learn more at Retool.com slash agents. Again, Retool.com slash agents. Yo other NoSQLs, your tech is a wreck like an oil spill. You call that web scale more like toy scale your MIM map file is a database fail and your users gonna set

sail for the one they can trust cuz if you're super yust then you must be running couch base Wow Wow yeah man that turned out to be true it's nice when you put together a rap and it turns out to be true as well, you know? It's like a bonus. Yeah, I feel like it's been literally 15 years though since that moment. Almost 15 years, right?

15 years of focusing on the same thing in a lot of ways and what we're doing with vibes is an outgrowth of that. What does that focus? If you could just put a name on it, what are you focusing on? Programming is way too hard. It shouldn't be a thing. Like if you're me, code is never the hard part

and I want it to be that way for everybody. And how are you doing that 15 years ago with CouchDB and Couchbase? Well, my role over there was the CouchApp side of the project, which was to make CouchDB into this serverless application runtime, you could just drop some JavaScript into and go. Did you get that done?

It was, um, it kind of still is responsible for a lot of people who otherwise wouldn't have had infrastructure to support them being able to do things like rural healthcare or fun apps for their club. A lot of the same things that we're seeing people do with Vibes DIY. So for our listener catching up, Chris was first on the show episode 54 back in 2011. CouchDB, which is, I assume it's still a stand,

so you have to forgive me, Chris, I'm not out there using it, but was a massively popular open source database, businesses formed around it. Damien Cass was the original creator or was it a group of people? I mean, there's a lot of people on the Wikipedia page, but help us understand the culture there.

Damien's the guy who CouchDB came to in a dream essentially. He had been working at Lotus Notes and getting frustrated with the fact that it had to pause the world to do its synchronization. So he invented the data structure that allows the streaming synchronization with fairness. And how'd you get involved? I started using it to spider the web.

This was like early, early on and had to teach myself Erlang essentially so I could drop a debug log function in there for myself. Next thing I know, I'm through the whole stack and never looked back. Nice. So what'd you do? Reach out to Damien and say, hey, I'd love to help out

or let's start a business or how'd that whole thing come together? All good things start on IRC. That's a great answer. We were hanging out on IRC. I think it was Jan Leonard who came out to Portland for OSCON like we were talking about. You know, we're already buddies from online.

And then after that, Damien invited the brain trust out to his place in Asheville, North Carolina, where I said, hey, let's start a company. So my role in that besides not being scared of the code was not being scared of the lawyers. Tell the story of that company from there. Tell the story. So as long or as short as you want it to be. There's a lot that went on,

but the thing that really sticks with me is the fact that we kind of accidentally incubated Node.js. We had people like Isaac Schluter and Ryan Dahl hanging around the office, mostly because of Michael Rogers, rest in peace. And it was that crowd that was just informally writing NPM at the desk next to me. So really fun to be in the presence of greatness

for so long and to be able to incubate things like PouchDB, et cetera, kind of taking JavaScript to the next level. My first exposure to CouchDB was via a, now we're name dropping, a Jeffrey Grozenbach peep code on CouchDB was via a, now we're name dropping, a Jeffrey Grozenbach peep code on CouchDB. And he made it look so cool. I was just a fledgling Ruby on Rails developer,

of course, dipping my toes into the JavaScript ecosystem, and kind of exploring databases at the time, and like wondering, because Rails is so easy to kind of switch your database adapter, but nobody's really doing that. But I was interested in the different databases and like CouchDB was so radically different than what I've been using.

And yeah, Jeffrey's voice, which is epic to this day. Legend. Yeah, he's a legend. Made CouchDB look like this is where the cool people are coding. And it turns out it was true. I mean, from your story is like, yeah, a lot of very talented people are involved in the project.

You have to remember 2008, 2009, that was the very beginning of the API revolution. The idea that you could put your database behind a REST API, almost nothing was even JSON yet. So we were all frustrated with XML. We were all writing kind of monolithic servers. And by the time that project kind of matured even as early as 2011, at least the early adopters

were already doing microservices. And so it was really, that was a big sea change right then. Mm-hmm. Was it good business too? I mean, you started a business around it. Was the business good? Was it hard? How'd that go?

It was good and hard. And in some ways we brought like a cultural cache that allowed us to, as we discussed, like on episode 54, merge with Membase who had a whole bunch of really smart people who understood the market and how to drive a startup to success. So, you know, Steve and me,

who I learned revenue from over there, is now on the board of Vibes DIY and helping me do the right things here. It took a while, but they IPO'd. They IPO'd about five years after I left. And, you know, always watching it real close. So what's your journey from leaving there to Vibes, Vibes.diy, which is your current thing?

Well, I bounced around a little bit, you know, checking out other database technologies, worked on FaunaDB, which I still maintain has some of the strongest integrity that you can get. Then I went to McKinsey and Company, kind of a fun story there, like you wouldn't think that. Oh yeah, my whole family was like, really?

And the, but you know, Jason Smith is another legend. He's, he's still there, but he brought me in and his claim to fame is, well, for Couch, he hosted NPM. So NPM used to have a one-man host and it was him. And yes, so great to have him show me what the real world is like. And what I learned about the real world is you don't get infrastructure, right? Us here in the Silicon Valley culture, where we can, you know, boot a virtual machine or request an S3 bucket and get it,

like that's just an unbelievable luxury. What I learned from this constrained environment of the client doesn't know what cloud you're gonna use until it's too late, is I need to build a database technology that can run anywhere to allow permissionless innovation. And so that's the gem of what became Vibes DIY. I built a deep tech thing along with Michael Rogers team at Protocol Labs who did a whole

bunch of research on immutable data structures with content addressing. I joined that team for a little while and saw the potential of it and built an embedded database that runs in the browser with end-to-end encryption for multi-user synchronization. And that's what powers Vibes DIY. The database is called Fireproof. Basically everything I'll talk about today is open source.

But if you wanna get into FireProof, there's plenty about it. The thing that it does uniquely is give every operation cryptographic data provenance. It's almost like a mini blockchain in the browser, super lightweight, no network dependency, just runs right there. But it means that when you do a replication,

you know what you've got. There's not questions about what kind of data you're sitting on. Let's dive into that. When you say cryptographic data provenance, what do you mean exactly? So it goes back to rich, hickey closure and the immutable data structures that they were pioneering around the same time we were doing Couch. And what you see in those data structures, as soon as you have an invariant that you're working with immutable backing storage,

you have to do everything different, everything new. With Fireproof and the IPLD data structures that it uses, that constraint is enforced because the only pointers you're allowed to use are hashes of content. And so everything is addressed by its own hash. That means you can only talk about stuff that's already been written. And that means that every time you're

doing these B-tree-like updates, you're either going to churn data in a real heavy way with lots of right amplification that's unnecessary. Or you're gonna solve these hard, hard problems that the research team over there, people like Alan Shaw and Volker Misch worked on, where you have trees that have really interesting properties.

So what you need and which is a lot of work to get is something that is a search tree like a B-tree, but regardless of if you put all those records in in one order or a random order, you still get the same root hash at the end. So your snapshots have stable identifiers. And that means your replication doesn't have to be mechanically the same everywhere.

It means you can replicate essentially in an optimized efficient way, but still end up at the same end point. So fire, fireproof, firestore, what's it called? Fireproof. Is this your implementation of these concepts or? It's my essentially making it so easy to use. We talk about technology, especially databases. And one of the phrases people like to say is a foot gun. A foot gun is when if you don't do everything exactly right, it blows up on you. And the requirement I had for this technology

is no foot guns. It's literally designed for the code school dropout Okay So you can't actually shoot yourself in the foot Right and that turned out to be the right place to be because we got a lot of people vibe coding Well, even more than that I'd say with LLMs and GPTs and all that we have an infinite supply of code school dropouts. Oh true Yeah, exactly. Where they are infinitely supplied

depending on how much money you have, right? There's another one there as assuming you can afford those tokens. So you're working on Fireproof, you're building this, making it simple for people to build things with it. And then where does Vibes DIY come into the story then? So we were going to market with Fireproof after the technology got mature.

And there's use cases like black box recorder for self-driving car. Like things that are very serious that are served well by this content addressed immutable data structure. But there's not very many people who need that and they don't care how easy it is, right? That's a different set of problems for them. When we started to see in our Discord people coming in saying, I haven't programmed in 15 years, but I built this drum machine along with chat GPT and your API,

and it worked right out of the box. We realized we've got vibe coders. And so I brought in an old friend of mine, this guy Marcus Estes, who I've known for like 20 years, and he's just a hustler, right? He makes it happen. He understands where the opportunity is and how to go after it. I've been looking at this vibe coding stuff thinking,

that's fun, that's cute. He realized like, no, let's just do this. And we didn't look back. So it's been so much fun. The big difference, like the real difference that matters is my kids who were so sick of hearing me talk about Merkle CRDTs and all this stuff. Now they're grabbing my phone out of my hand, right?

That's cool again, you know? They just wanna make apps. They're like, dad, can I make an app? And you know. Dad, can I make another Merkle tree? You know, whoever says that. Right. What kind of apps are being built with this thing or can be?

Well, I was really inspired. I was trying to see, you know, essentially what's the lowest common denominator that we can address. And the answer to that is, I was using ChatDBT Canvas and Cloud Artifacts and thinking about how for like a normie, somebody who doesn't already know their way around

the next JS stack or deeply understand React and all the tool chain there, those are the best kind of interface and set of metaphors. If you look at those closely, they don't even say deploy, they say publish. And so that's the take we've done with Vibes DIY is if when I'm designing a feature, if it's gonna require introducing a new concept, don't.

Just make it simpler. And we had this tremendous validation. We just got back from Render ATL. It's a tech conference in Atlanta where we had a deep line at the booth crowded up with people who had kind of maybe never even heard of VibeCoding. Or if they had, they hadn't tried it yet,

or if they had tried it, they thought the existing tools were too complicated. And when we said, hey, if you make an app, we'll install it on this sticker for you, they would spend the 90 seconds it takes to make an app. So. Wow. What's on the sticker? Is it like a QR code or a install kind of thing or launch it?

The way we install it in here is with the near field communications chip. So. Okay. I actually learned about that at like a Burning Man event where someone was doing manicures with NFC chips in them. Really? Yeah, which is super hot.

Like, here's my number. Right, like you can just embed a little bit of information on that chip and then carry it around in your fingernail. Yeah. That sounds very burning, man. Yeah, you learn a lot. You never know where you're gonna learn a lot. Here's your next big idea.

Right? Wow, so what, so if I take the sticker home, how can I use it? Like I got this information embedded in it, what do I do? You just kind of get it close enough to your phone and you get one of these little, and now you can open the URL. Okay, so similar to a QR code,

except for not for the camera, but actually the NFC chip in the phone, makes sense. That's interesting. Right, and for us, it's really this ownership metaphor. Like code is all in there, over there, especially if you're not a coder. Like, what am I making, where is it? But if I put it in your hand,

when you walk away from the booth, then you're gonna come back with your friend the next day. Does that give them access to the code they just created or vibed? Or does that just give them access to the code they just created or vibed? Or does that just give them access to the application they just vibed? Well for us really our... Like continued development, how they keep going. You're totally right.

So most of these people made it on their phone they walked up to the booth with because it's a mobile first experience. But if you didn't, if you made it on my phone or if you made it on the booth laptop, then you still get the sticker, you can still scan it into your phone. And anyone who goes to a Vibe can hit the remix button in the corner and get into the code or get into the chat and say,

the default remix we have preloaded in our forum is make it pink, but really just whatever you wanna do to change it. Change, there's a great story, we had a lot of fun. I was walking back from lunch in Atlanta and I said to my phone, I said, I wanna take pictures and have you turn all the faces in the pictures into Georgia peaches.

And it worked on the first try, you know, sometimes you get lucky. And people thought that was hilarious. So we were showing it off and then, you know, we were sharing a booth with Netlify. They're great friends of ours. And one of the Netlify leaders, Sean brought his kid.

And so she saw what we were doing and she said, I wanna be a cat. Right, so he next it. Basically a four year old turned the Peach app she saw what we were doing and she said, I want to be a cat. So next day, basically a four year old turned the peach app into the cat app. And that was a hit. And then we saw what was going on there and we made this. We made a caricature app I've got on my laptop here. Here's me as a podcaster. And, you know, and it was it was also a big hit.

People at the conference just all day long building stuff. Just fun. Wow. That's the stuff I'm talking about, Adam. Just total vibes. For joy. This is the internet, man. It's for the joy of it.

It totally is for the joy. And so many people. This is like GeoCities for the modern era. Seriously. It's like a writable GeoCities, right? View source is back, baby. Yeah. I'm reminded of the old Cloud to Bud extension.

You guys remember that one? Every time the word cloud was on our website, this Chrome extension would replace it with the word bud. I'm not saying this is an idea for a remix, Chris. I'm just saying I'm reminded of it. You know, I should, if I, if I didn't want to get distracted, we'll just build it right now.

Yeah. We'll just build it right now. Well, friends, it's all about faster builds. Teams with faster builds ship faster and win over the competition. It's just science. And I'm here with Kyle Galbraith, co-founder and CEO of Depot. Okay, so Kyle, based on the premise that most teams want faster builds, that's probably a truth. If they're using CI providers

for their stock configuration or GitHub actions, are they wrong? Are they not getting the fastest builds possible? I would take it a step further and say if you're using any CI provider with just the basic things that they give you, which is if you think about a CI provider, it is in essence a lowest common denominator generic VM. And then you're left to your own devices to essentially configure that VM and configure

your build pipeline, effectively pushing down to you, the developer, the responsibility of optimizing and making those builds fast. Making them fast, making them secure, making them cost effective, like all pushed down to you. The problem with modern day CI providers is there's still a set of features and a set of capabilities that a CI provider could give a developer that makes their builds more performant out of the box, makes the builds more cost effective out of the box and more secure out of the box. I think a lot of folks adopt GitHub Actions for its ease of implementation and being close to where their source code already lives inside of GitHub. And they do care about build performance and they do put in the work to optimize those builds. But fundamentally, CI providers today don't prioritize performance.

Performance is not a top level entity inside of generic CI providers. Yes. OK, friends, save your time. Get faster builds with Depot, Docker builds, faster get up action runners and distributed remote caching for Bazel, Go, Gradle, Turbo repo and more. Depot is on a mission to give you back your dev time and help you get faster build times with a one line code change. Learn more at depo.dev. Get started with a seven day free trial.

No credit card required. Again, depo.dev. So is the platform the web then, or how are they how are these things running? That's the way I think about it. I wanted to build the experience I had early on in you know the black and white Mac days with hypercard but I realized that when I look around people trying to build similar ambitions, often there's a do it our way completely or don't kind of feel, right? I call that like they make you wear the socks.

With us, the point is there's already so much platform that's done. So we just build React apps and we're essentially taking the web runtime, right? Which is like what a browser can do and making it as easy to express yourself there maybe easier than making a TikTok. So if making an app is easier than making a TikTok, now what? And that's the question VybzDIY is out to answer.

I love it. I think Adam and I were just talking about this new possibility, specifically in light of Apple allowing local models on phones and Xcode and like things are getting localized there where it's like you could build and deploy your own iPhone app.

It's just a one-off just for you. Like nobody else even has to have it. And there are still like platform limitations there with regard to distribution, side loading, and there's lots of stuff. But with the web, it's like, nah, just URL, publish that sucker. And that's really awesome.

So just to nerd out again on the sticker, is the embedded information in the sticker just a URL to where the thing actually runs, or is there code in there, or how does that work? It's just the URL. So once you have the URL, we make, it's actually our service layer, so simple. We have your, so if you look at a lot of the existing

vibe coding tools, they're built for professionals and they're gonna like mimic a professional output with a full Next.js app. The only thing that the model, which is just a commodity model, we have a switcher. It's actually a great way for AI interested devs to try out all the different models and get their personalities down.

But when you ask Vibes to make something for you, it's only writing app.jsx and that's it. So our focus is on being fun, done, and alive with AI. This last bit means that the apps that you make get the same API key that we use in your browser to write the code. And you can press the demo data button and it fills out your to-do list

with whatever is plausible. Or it makes an app, I built this because it takes 90 seconds to make an app. It takes longer to figure out how to use an app. So built this because it takes 90 seconds to make an app, it takes longer to figure out how to use an app, so instead I just mash the demo data button and it shows me what the app would look like if you'd spent some time in it. So let's say I mash out a 90 second thing

and it's rad and we're having fun, can I then expand, extend, like build that sucker out or do I have to remix and start something else? That's where one of our core values comes in is that it has to be real. So the 500 lines of JSX that it's going to write is React code. It's as good as Claude 4 is going to make it, which means it's not that bad. And the standard library we bring to the table with it means that all these apps are fully functional just in the browser, but then they can be remote synced via

the cloud to other users for collaboration. And that means copy that. There's a copy button. You get your 500 lines of code in your clipboard and you go drop it into anywhere and maybe into a Vite template or maybe into another vibe coding tool and say, hey, expand this. Maybe you turn it into 10 pages. Maybe you take 10 vibes and put them into one app. But there's, the sky's the limit.

It's really fully malleable software that pros can take and run with as well as beginners can just use. Adam, what do you think of all these sayings he has? He's got like rhyming words, he's got like stickers. He's a poet, man. This guy's bringing it. I'm over here trying to vibe over here.

I think it might be a safari issue. I thought you might be, that's why I pulled you back in. I'm like, I can tell when Adam's like testing something out. I was vibing over here, and I think I might've hit an issue with safari potentially, cause the thing it made me is on a black screen, and so I'm having an issue. Well that's not safari. That's Claude.

I would say you rolled the dice and one. You got a snake eyes. Oh, dang. One in 20 times. And one in 20 times Claude is like, oh, I know what you want. You want a black canvas that's like 20 times bigger than your screen.

It just does that. There's, we have work to do, but. That's about, yeah, I think that's about it. It's kind of like forever scrolling, so. Yeah, I mean. Well, close the tab, open a new tab, and try again. I can share with you my, it was just intense, man. This is an intense process.

I saw it write a bunch of code, I could see all the code, but there's no application to be viewed. Basically copy that prompt you did, just make a new one. That's what we learned from vibe coding. Don't go iterate on something unless you're already stoked about it. All right. It's just cheaper to throw it all away

and start fresh, right? Yeah, right. I mean, when it's just spoken word, it's like, let's try again. And just tweak a little bit, you know? That's kind of what I do. Remember with Google searches, remember when you used to Google things?

Some people will put in a search and then they'll read the results. And if there's not the right answer, they'll go to page two, page three, page four, like an infidel. I just research, like, oh, my search wasn't good enough. I just research. Now I have the same thing with these LLMs.

I just copy prompt into four of them. And I'm like, let's see who's got the best answer or the fastest or whatever. Or I can like mentally munch them into one answer and like pull things from each one. And that's kind of the new research is just like, nah, I'm not gonna like tweak this prompt. I'm just gonna try it on more things and see what happens.

So yeah, close tab, open new tab, paste prompt and start vibing again. Yeah, we're considering seriously doing that for you. We may as well just vibe three of them in parallel and let you pick the best. Although every bit of complexity we add is gonna take the simplest user and turn them away. And cost.

I mean, that's what Sora does, which is OpenAI's video tool. You can pick how many different iterations that you want and it will make two or four or six. Obviously you have to have the tokens and stuff. But especially with creative tools where it's like, I would like to see four different things. It's just like, yeah, we'll just crank out four versus you having to say, you know,

try again, try again, try again. The fun part is looking at how much better Clawed 4 is than even 3.7, or if I switch it over to Codex Mini, then I'll get like an app that works great, but isn't showboating at all. Like GPT models make these lean apps with no extra stuff,

and Claude 4 is gonna have your like background, transparency with blur and movings. You know, it's just, it's fun to see the personality in there. I think Claude Four, and I don't have any personal reason to say this in terms of like, we're not making money by promoting Anthropi, but Claude Four is just,

I think good enough now that it's got, it's sold me on the entire thing. Whereas up until now, I've been very skeptical because the results have been so bad. I'm not sure if you guys saw Simon Wilson's the last 12 months of LLM progress as outlined by Pelicans riding on bicycles. You guys see that post where he basically every because he's such a nerd for those for every new model that comes out. He tells them all to do the exact same thing, which is to write a code that produces an SVG drawing of a Pelican riding on a bicycle.

And he documents that, of course, because he documents everything. And he's just showed the progress. He was like, here's this model, here's that model, all the way down through the last year, and how far they've come and which ones do better. And he scores them and stuff like that. But the progress has been amazing.

And Claude Four, I think we're at a point now where it's like above my threshold of sucking to where it's like, let's do it. Like I wanna use them now. I'm not sure what happened between three, seven and four, but something happened. That black screen that you ran into Adam, the other one that I'm just still trying to,

you know, teach Claude to never do again. For some reason it loves to give you buttons with white text on white background. Dark patterns, the opposite of dark patterns, light patterns. Did you get better results this time Adam or are you still? It's still working on it.

It's a little slow for some reason. It's taking its time. It's bleeding edge. How many lines of code does it wanna do? At this point it trying to again. So I got two black screens in a row, so I wasn't sure. What are you trying to build? Okay. So I told this thing, I want to golf with friends,

help me create a tee time and invite folks to it. So, Hey Jared, I'm going to golf this Saturday at this time. You're invited to my tee time and you can join. Kind of thing. So that's, oh, it finally did something. It's got a golf tea time organizer. I can create a new tea time. I can invite people to it. It seems it looks like more iteration.

It's got some additional features that could be added, which I think is kind of cool, but all in all, quite colorful. It's it's vibrant. It's got a cool vibe. It's vibin'. This is vibin' man, this is vibin' man. So do you guys like have a system prompter? I'm sure you have some sort of scaffolding in there

that when we write our prompts and put it into Vibes, it does more than just that, right? The whole thing's open source, so you can go on GitHub on the vibes DIY org and get into the repo prompts.ts and see what we actually do. But it's fairly simple. I mean, as I was mentioning earlier, I was one of the co-authors

on the O'Reilly Chat GPT Shortcuts book. And what I learned from that and the experience continues to pan out. If you talk to the LLM like a person, like less technical, more just what you want, you get better answers. So this golf thing is the thing you wanna do. There's this great YouTube,

we can find it for the show notes. There's two types of vibe coders, and there's the one guy who's all frustrated with his 4,000 line LLMs text and his feature specification PRD and the other guy who's just like make GTA make GTA six and he's having a bright you know here's my golf tracker oh nice did you do that one just now as Adam was doing one or is that yeah I heard what he was doing. And I think my prompt, my prompt was just. What is it again?

Show it to me. I said golf with friends. Oh. That's all you said? Yep. Oh my gosh. Now, okay, so how do you share that with me so I can see it or play with it or remix and revive?

Pass the address into the chat. We have a chat here. Right. So we don't deploy. You press the purple button, the big button, and now you publish it. Publish. I was getting to this a minute ago.

I guess I distracted myself. So when it publishes, all it does is take, like let's say the 475 lines of JSX and put it in a Cloudflare KV associated with the subdomain name. When you go hit that subdomain, we have a static HTML that's essentially the runtime that's got all the packages in there and it loads any additional packages via esm.sh. And now you have your app right there, costs us almost nothing to host, almost nothing

to serve, unless that app goes and does some AI calls, in which case that goes against your token quota. So it makes it super easy for people to write these apps that are alive with AI, whether that's maybe our benchmark. Compared to Simon Willison, he's got the penguins on bicycles. What we have is, for instance, I'll just tell you something in English and you'll give me a structured to-do list that reflects back what I need to do to accomplish that.

So I'm always running that when we do an upgrade or move to a new model. And this morning I said, take my kid to summer camp. And it gave me everything, sunscreen, lunch, all that. All right, I just loaded up Adam's golf tea time organizer. Is it working? I mean, it rendered. I right, I just loaded up Adam's Golf Tea Time Organizer. Is it working? I mean, it rendered.

I don't know if it's actually working. I think it has a white button with white text as Chris. Something's going wrong here because I can't get the load for me. But I can create a new Tea Time. Oh, I see some. Sunnyvale Golf and Club. I can join that round.

Yeah, pop your name in there. Boom. Okay, I'm joining that round. My name is Jared. Thank you very much. And confirmed, Jordan Phillips is going and Jared is going. So where is it storing that information, Chris? Where does like the data that I enter in go?

So all that stuff goes into Fireproof, which has got that Merkle CRDT. And that means that it just goes in your browser local storage. The way Fireproof works is every database operation is a diff like in Git that is end to end encrypted and then written to, we use index DB in the browser because it's real fast with Uint8 arrays, but then asynchronously can be replicated via S3 bucket and WebSocket. So that part works in Fireproof.

We're rolling it out in Vibes right now. That means today all Vibes are single player. That's why it's best for these kind of viral AI experiences. But tomorrow, all Vibes have essentially the security model of a Google Doc. So I don't want an end user to have to understand like Postgres row level security or having to do like where clauses to keep from writing bugs.

Instead, if you're in the app, you're in the app. Maybe you have write access or read access, but it's as simple as a Google Doc for the sharing. And that means, you know, as soon as we activate that feature, then, Adam, you could invite Jared to your app via email. And now y'all are both in there setting up what are you going to bring for lunch to the course.

And it's all on the web platform. Yeah, yeah. None of the people who don't know what code is, who've never deployed anything, they can't depend on any, you know, setup. I like to say, if you don't know what an API is, then I'm certainly not gonna ask you to do an API key. So this thing went render ATL viral last week

because they already had a fun time making caricatures of themselves, but when's it gonna go, you know, TikTok viral or youth viral? Like it seems like this is set up specifically for 13 to 15 year olds to find out about it. Just like the six, seven thing. I'm not sure if you guys are aware of six, seven, but they're huge on six, seven.

And all of a sudden it's like this thing's screaming. Well, I'm pretty sure that my key into that market is you guys. What we aim to be like there's a lot of people who use more serious tools for vibe coding and I'm not gonna try and be you know I'm not gonna give a senior developer more control more options more details than other incumbents but what I will do is be exactly the right thing that they're gonna turn around and show their friends

when they want their less technical friends to start getting into writing code. So if you've got a buddy who has a dream or keeps bugging you to build something for them, say, hey, vibe it up and I'll take it from there. Right. Well, it's very much the GeoCities then in that sense. Or maybe Glitch.

Of course Glitch recently changed the way Glitch works, but similar vibes then, right? Glitch was very much a thing where it's like get people started, you know, learn on it. Yeah. How does versioning work in that multiplayer world? So if you look inside the chat that you've been having, you can refine it.

You could say like, oh, you know, I didn't mean regular golf. I have these interesting rules of golf I wanna bring to the game. Claude will go rewrite the code for that. Now you have a slightly different version of the app. That could be within a single chat session, in which case we're not remixing yet.

We're just iterating. But if you go look in that chat, you could click on the older versions of the code and it brings them up instead Whichever one is on your screen is the one that gets published when you click publish So publish an old version if the new one wasn't as good But once it's published and you go to that URL, you know When you scan the sticker and you go to the URL

Then you're gonna have a remix button in the corner of the screen and when you hit that you've essentially to have a remix button in the corner of the screen. And when you hit that, you've essentially forked. Now, no insult to all of the people out there who are kind of like me in this regard, but I've been forking forever. And what we say to ourselves to make sure we stay on target, Vibe's DIY, is we say forks are for dorks. Well, this guy's got a saying for everything.

So because you don't like forks or because you do, I mean, it depends on if you think a dork is a good thing or a bad thing. Right. Well, it's time to get the dorks involved, do the forking, but when you just want to try something and remix, right, then you don't need all that conceptual baggage. Just hit the remix button and go. So this is akin to the publish button versus the deploy, right? Like, it's like remix it, don't fork it,

because you're not a dork, you're a normie, as you said earlier. Yep, yep. Well, you know your audience, and it's not us. It's not us. Because I'm looking for the fork button over here. I'm like, excuse me, sir, I would like to fork this. Totally.

Well, it is us though, because when you go deep on something, it's got that Merkle CRDT, it's made out of standard React components, it's got all the TypeScript annotations you could hope for, and it's ready to go big if that's what you wanna do with it. But I imagine sitting on CloudFlare with an HTML file and some KV backends that big in terms of maybe features

and commerce and production big, business big, but not big in terms of scale. I imagine one of these could scale pretty easily in terms of that TikTok viral moment, couldn't it? Cost pennies on the dollar compared to anything that requires a VM or any kind of complex state or CPU. When I think of Kubernetes, containers, all that, I only want one if I basically want to rent some live RAM,

and you don't need that. All the live RAM lives in the user agent at the edge. Right. So Cloudflare is pretty cool. I mean, just, no further comment. I wonder if you could maybe expand on that. Like you guys are doing cool stuff because of what Cloudflare is helping you to do, right?

I mean, that's just, it seems like a very interesting platform in that way. Our stack is, I guess, native to that way of thinking about things. So not just the publish flow that goes into a KV, but also the fireproof synchronization is powered by Cloudflare workers, which is a great way to rent some RAM on the cheap if all you need is a list of your currently connected clients.

So essentially the flow is that tiny encrypted diff, you write it locally, push it up via our API, which you don't have to know about, but it ends up as another record in an R2 bucket and then an update to the pointer that corresponds to your database. So everyone who's listening to that pointer pulls down the latest. And because it's a CRDT, the merges are resilient, right? I was thinking about recovery mechanisms. Say like you lost the pointer and all you had was the files. It's a valid exercise to just load them all and merge them.

And you'll get the same result as if you started from the end one. Why RAM? Why does it need to be in RAM? I mean, it doesn't. I just, we can do it with S3 bucket only. It's cool, right? It's a little faster to keep it on S.

Web socket. And what I mean by in-ram is- Is it necessary though, or is it just because it's cool? It's just because it's faster. So we need essentially a directory, a link to, okay, let's get technical. The real reason we need any of that is because it's multi-writer safe.

And so if you just had like a SQ light and you were trying to do this kind of architecture, which people do where they put the SQ light segments into an S3 bucket and then they load them in the browser and query them, that's all good until you need to start writing. And now you're doing some kind of bespoke, you know, take the write ahead log and put it in there and manage, you know, kind of conflicts against it and everything. It's not for the faint of heart. Instead, what we do because of this Merkle CRDT, the merges are handled for you.

So the only thing we need to add to the recipe to make it multi-writer safe is it could be that pointer that points to the most recent file, the diff from which you got to start to reconstruct. Well, it can point to more than one file. So if you're under concurrent load, people can all, you have 50 people all right on top of the same starting state

without having coordinated each other yet. And now we just have a pointer that's 50 wide, the next read pulls those in, merges and writes back up the single pointer. So that's what we do the performance thing for. So in your mind or in your heart, is fireproof the thing and vibes DIY is a thing to demonstrate the thing,

or is vibes DIY the thing that fireproof lets you build? I feel like I'm in the Millennium Falcon and I just like mashed the gas you know so it's a little bit of both but it's it's super aligned. Fireproof is for the kind of people who like to fork stuff and right don't want to be bothered with all that it's just done. Okay do you think you're out to pick a certain point or like focus or do you have infinite bandwidth to just like mash that, mash that millennial Falcon?

What's it called? Warp speed? No, they don't call it warp speed. What do they call it in Star Wars? I'm blanking. The hyper drive? Yeah, the hyper drive. Go to hyper speed.

So we are focused 100% on vibes. It's, and in a lot of ways it's because Fireproof is basically done. I mean, it never is the case with a database, but it's not like it needs new features. It's just about hardening it and continuing to release it. So in that world now, Vibes gets cool new features, like a fun new feature we get to add.

This is lightweight stuff. It's like, like I said, I don't want people to have to know what an API is, much less an API key. So bake that YouTube API key in there. It's not in there yet, but this allows you to say, hey, I want to turn my playlist into a YouTube screen and it just works already. Same thing for Spotify or any other of these kind of mass market media APIs. There's no reason that you as a VibeCoder should have

to know about any of that to use it. Have we talked about money yet? No. How do you make money? No. Let's talk about that. Let's talk about it. Does it matter at this point or is it just for the vibes?

It matters that we don't get it wrong and then the more we think about it, the more excited we get. So if you look at what it would take to build something with like a vibe coding tool and a database backend, those are both kind of like maybe just to get started, table stakes, $25 a month, $25 a month, that's $50 a month right there. We do an all in one bundle at $5 a month.

And that's because all we need is that R2 bucket. It's not costing us a whole Kubernetes container to keep your Postgres instance running. Makes pennies on the dollar for us and then the real cost becomes those tokens that either your app consumes when it's generating cool caricature images or parsing your to-do list as well as the tokens that you use when you write an app. $5 a month gets you started. If you want to go over that, there's a meter, which is competitively priced. And then where we see this going in the long run is almost like this YouTube style model where say, you know, first think about a corporate case. Think about some big marketing agency that builds like, see our product in your house.

Replace the Statue of Liberty with our product, whatever it is, right? Take a photo, put us in there. You get 100,000 views on that from your email list. You're probably writing us a check for $20,000 or something on the tokens it consumes. But if you're some kid who figures out a cool way to scan your house and turn it into a video game level,

you don't have that kind of money, but your app costs just as much to run and it goes just as viral. So in that world, there's a little bit of a, you know, freemium, somebody comes along, uses the app, they can generate a simple level or a couple images, but then they're going to get through their set of free vibes tokens. If they like what they're experiencing, then now they're subscribing for $5 a month. So we're not just selling

it to developers, we're selling it to people who want to make cool caricatures of themselves in their dream job or whatever it is that these end user apps do. And so now if you're that kid who made the cool video game and you don't have any money, but you get us 50,000 signups, then my goal is to write you a check like YouTube. Does that turn your house into a video game level thing exist? Cause I want that.

No, somebody got to make it. I mean, why wouldn't it exist? That's such a good idea. Right, Gaussian splat is cheap now, run it on the phone. How would you do that though? Would you just walk around your house just like taking a video and uploading it? How would you?

We like to say that the best we can hope to do, you know, we're kind of grizzled old vets of the industry. The best we can do is invent the electric guitar. It's up for the kids to invent rock and roll. That's like the best non-answer ever, right? He's like, you guys figure it out. All right. Yeah, I mean, we've got roadmap features that expand.

So everything I've talked about is really just the browser runtime. It can do a lot, it's underutilized, especially for people who don't really, you know, haven't pushed it to its limits yet. But roadmap features is like something we learned from the CouchDB ecosystem. Essentially Fireproof is the CouchDB model

running in the browser. And one of the things that you see a lot of, the way NPM was powered, all those packages stored in CouchDB, and then when someone uploads a new one or makes a change to some metadata or whatever, there's just an event reactor subscribed to the database feed that can then go do some heavy lift or send an email or update some analytics package somewhere.

Same thing, just if you wanted to today, you could stand up a Cloudflare worker that subscribes to that golf database. And when somebody puts in a new tee time, you get a push notification. Our roadmap brings that to the masses by making it so there's just a backend.js that we also let the model write. Yeah, so I wouldn't even have thought about that, I guess. So it'd be helpful if there was a guide, you know, like here's the next step you can take. They could mention different features can come,

but not what you mentioned there, which is like push notifications. That'd be kind of cool. Right, and there's no end to that kind of stuff. And we really, essentially, if I think about the job you're hiring that backend.js to do, it's the stuff that, you know, essentially hiding API secrets.

You can't just publish the way that you send your emails, your postmark key to the world. You can't put it in the browser, but you can put it on the backend, and then the backend's job is just to forward those messages. What happens when it's successful? What is a year, two years from now, how does this change things?

What is vibes.diy a year and a half from now? Well, now that it's easier to make an app than a TikTok, we picture people, you know, swiping right on the good apps and long pressing to launch it. And, you know, having essentially, you know, their phone blow up in their hand when their app gets popular.

So the same thing you see with an Insta or a TikTok or a YouTube where it's really about the excitement you get from coding. That's what I saw in Atlanta. When I was first starting coding, right? Like the coolest thing in the world was just to change, you know, some one line in your HTML and then hit refresh and there it is. So educating people how that feels and then, you know, letting them share

that with their friends. I think what we see is that coding becomes something that is cool. What do you how do you deal with the potential of scams or just like security issues in that unfettered world where it's like, yeah, check out my Vibe app and it's cool, but like what happens when it's a scam? There's thankfully a whole bunch of prior art. Essentially for us dev tool builders, it's a little bit of alien land, but if you've done social media before, it's that playbook.

So we've done some of the basic things already. For instance, there's no way to publish to a fire hose. It's not like just everything's all showing up. We moderate the stuff that goes out on the homepage. That doesn't mean you can't share something. And so if you start building honey pots, then now we have a different moderation challenge. It's not new. It's just new to do it with HTML instead of video or images. Yeah, I just wanna like, it's one of those things where you don't wanna have to deal with it, but you do.

And it's early, cause like trust will be everything to get to the TikTok stage of sharing and virality or whatever, it's gonna be something you can trust or has to be. And if you can't trust it, then it's like, it's a today something you can trust, or has to be, and if you can't trust it, then it's like, it's a today problem, basically. Like at launch. Yeah.

So, if you look at the history of using mechanical Turk to moderate profile images, and having to do two of three thumbs down and all that, so much cheaper, so much easier to do that with a model now. Just let Claude tell you whether it's not any good. And then on top of that, so much cheaper, so much easier to do that with a model now. Just let Claude tell you whether it's not any good, you know? And then on top of that, we aren't gonna run into these problems, it's sort of blunted. We don't give you the ability to go into the editor and type.

That wouldn't be hard for us to, but we figure that's the point you should copy and paste it somewhere else. And Claude's not gonna write something out of the box that's gonna be extremely concerning we Just use the open AI image gen. It makes our job easier eventually, we'll want to value engineer all that stuff, but That comes with scale Solve it with more AI. That's interesting right

With more AI, that's interesting, right? All problems get easier with AI? Well, AI makes new problems faster than it makes the other ones easier, but hopefully we've isolated ourselves from that with this one file app.jsx thing. And don't underestimate the power of the single origin policy, you know? The web browser kind of already worked out

a lot of this stuff. So if this thing takes off, I mean, now is the time to get in there and really be the vibe DIYer in chief, right? Like the people that were early on YouTube, the people that were early to TikTok, they had a much better time. Is there actual virality built into it?

Can I build an audience? Can I distribute? Is there a distribution channel? I know y'all are just getting started, so the network effects probably aren't there yet, but are the tools to build an audience for my apps in the product? So if we're talking 90 seconds from prompt to app,

you know, on the cases where it works, probably another 90 seconds to be like, is this app any good? And then hit publish. And now it's on a URL, so we go wherever URLs go. But the built-in features around like a newsfeed and a for you page and all that. These are the things I'm super excited to build.

We're growing the team right now. People who wanna do that should probably talk to me about it but it's definitely, it's a kind of interesting green field because so much of the design language is already worked out. It's not like we have to decide what an algorithmic feed even is. It's more just about making it for basically, we're bringing a new media object to the world and

Right the same thing you do with a video you do with this and we even have plans for instance The the same Jason Smith I mentioned earlier used to host NPM. I talked to him about it yesterday and he said, hey, I've been working on some video gen stuff. I think you could do, like we have this demo data button that fills the app up with data for you. We could also do a video tour button. You just click one button and it takes your face, lip syncs, puts it on TikTok, here's my app.

All right, that's cool. So what kind of engineers or people are you looking for? Really right now, if you go use Vibes, you're totally allowed to roast it because I am a React developer, I guess, but I'm also a CEO. And it's like 75% CEO code right now. So we've got-

That's C-level, yeah. C-level code, I like that. 5% CEO code right now. So we've got a C level. Yeah. See the little code. I like that. Uh, so if you're, if you're just like absolutely cracked and especially the person that the perfect hire is, you know, when somebody's kind of like new enough to a stack where they're like dangerously enthusiastic about getting everything. Right.

Uh, if you're senior enough, you're're gonna see my flubs and be frustrated, but at the right part of your skills curve, you're gonna just like chew up fixing that stuff and get it into best practices. Every level of CEO code is like an opportunity for the right person, you know? Yep. I just got like WordPress vibes in a way,

like in a positive way. When I was looking at this golf friends thing I got built out here, it's like like Kubrick like K2 was to WordPress very every app has a certain style kind of thing to it. I kind of feel like it's a good thing. Is there a shared is that on purpose Chris? Is there a shared design aesthetic or something? Well we made the choice to differentiate ourselves from everybody else we're not using Chad CN and that's not because it's not great. We just want to be different. We want to be more lightweight. We want to have a little bit more variance and

let Claude or whoever express themselves a little more in there. So the way we prompt the design language is actually my co-founder Marcus I mentioned earlier just did a bunch of research like art history type research around like we don't want synthwave that's played out etc etc we found there was a design school in Italy in the 80s that they came up with a style they called Memphis and so we tell it to make Memphis we describe Memphis a little bit and it does a pretty good job of that that's all user editable so if you go into the settings page right now, you can just type in your own style prompt,

you could switch it to Synthwave, you can switch it to Synthwave and DeepSeq, and then if you run it on DeepSeq now, you probably have to generate the app 10 times before you get a working one, but. Uh. Uh. Certainly a good playground for testing out these models

and comparing them to each other. Probably better than my copy paste prompt into like four things is actually making, probably better than Simon Wilson's Pelicans on bikes is like make the same app, you know, against all these different models. That's cool. I see the style prompt now is pretty cool.

You can just like swap that out, make it, make it whatever you want. Make it your own. Yeah. Right. Or if you really want to go deep, just hit up our GitHub, fork the whole thing, right? You know who you want. Make it your own. Yeah. Right, or if you really wanna go deep, just hit up our GitHub, fork the whole thing, right? You know who you are, and fix up my React. Yeah, I was gonna ask you that,

like how much of this can you run? You know, how much of this is just completely, if I had enough gumption, I could get it running on my own Cloudflare account. So, kind of yes, what we have set up is even for, if you're going in via the github it deploys to our endpoint and It uses our login system and that means that you could have projects

Basically, we want you to be able to be a vibes DIY user go run a fork Maybe your fork is like internal for your work or something But not have to do the whole heavy lift of the full stack. The fork will just run against our backend. If you want to get even heavier than that, yeah, now you can start to coat up the backend, but that's a senior engineer task. How do you keep up with your underlying technologies

in terms of what you're generating or what you're telling cloud to use? Cause those things are also moving targets. I have more work to do there, right? There's models that aren't represented yet. I would love to get O3. I use O3 personally when I'm in Windsurf. And so, you know, there's some things we can do to keep up.

We essentially need like a cultural ambassador to LLM land to keep us up. What about the next JS land? Is that a fast enough moving target? I guess you're always generating a new one. So it doesn't matter so much. Right. And we're just doing that one file. So I chose to do app.jsx instead of app.tsx,

mostly because it's one less thing the model can hallucinate wrong. Now, if you wanted it to be tsx, I'm pretty sure Windsurf can fix that for you in a single pass. Sure. But why? I guess, I guess by then you copy paste it.

So it's yours anyways. All right. Do what you want. Exactly. Yeah. The JSX is all about the simplicity and then the, um, it's really allows anybody to get in there, edit these apps and, you know, bring them to any backend. When you do eject, it runs through those same APIs. Again, you can move them over. I don't know if there'd be a good reason, but

there's a world in which, especially when we spin up the user pays mode for like image gen, AI usage by the app, you might just pull our NPM modules directly into whatever environment and use those because that lets you ship apps that haven't worried about the cost of the LLM. Say that again. I don't think I tracked it. So built into these NPM modules, it's part of our standard library is that when you run out of your free tokens, it pops up the login window. And after you log in and you get another batch of free tokens around those

it asks for your credit card if all you care about is going viral and the thing that's holding you back is that going viral is going to cost you $50,000 and chat gpt fees when people start using your prompts in production well do it on vibes or do it with the vibesibes NPM modules and you'll get the same benefit that will handle it for you. And if you get enough usage, then you can be looking at creator payouts. Yeah, that actually scales really well. I guess that's what you're referring back to the YouTube model.

Where you're basically as an app creator, I'm bringing you token use by way of people using my app and you being vibed DIY is making money off of every one of those people who decides to put down some cash in order to use the app some more and so you're now talking payouts to me for bringing you guys additional trap or additional customers that's that's an interesting model. A real kind of eye on the prize, you know, Couch did well, but I also learned firsthand what it's like to get whooped by MongoDB.

Hmm. How'd that feel? At the time, you know, I wrote that rap about other NoSQLs, right? You could tell how it felt, but now I appreciate what they did. They understood that the real answer to taking the market is to be on the ground floor when people are writing apps

without permission in the enterprise. And so we're probably a decent choice for say, you're running a veterinarian clinic and your you know Customer management system doesn't keep track of what the dogs favorite treats are Vibe up a treat tracker and you know on down the line the Cryptographic causal consistency that's enforced by fireproof means that you know, you could use this for regulated supply chain

maybe your cannabis dispensary needs to keep track of some stuff because you know someone with a badge is going to come knocking and Vibes has got the underlying infrastructure to make that safe. So after we make coding cool We have a lot of bandwidth to go and make this system scale for serious use cases. Do you think your brand scales that way? You know, we'll find out, but I kind of think so. Here's a sub brand, Serious Vibes. Right, totally. Yeah, we talk about gray mode. Gray mode. There you go. Yeah, there's a lot we can do there There's a whole like if you wanted to force the core thing and make it work with say like not

React like I got nothing against a view. I got nothing against solid.js. All these things are excellent technologies We chose react because it's the one that's gonna give you the deepest training set. But being able to take that, the wrapper and deploy a view app or a React app to our Cloudflare endpoint, it's not gonna change things. From an infrastructure standpoint, it's ready to go. All right, dorks, get out there and fork away. For the normal people, get out there and fork away for the normal people. Get out there and remix somebody else's golf teatime tracker.

Adams has already published it. And improve my vibes. My vibes were not so vibey. Get your friends into coding. Get some, you know, make an image generator that no one would have expected and make it go viral on your Facebook. I got a phrase for you.

Ready for this? I love friends. Vibe with friends. Vibe with friends, that's right. And that's coming, I mean, that's just an easy on switch. Some of the folks we've been talking to are most excited about that this chat interface where you're talking to the LLM

and it's writing the code for you, that's multi-user, right? As soon as we flip on the same sharing mechanism that allows Adam to invite Jared to the golf app, you can also invite Jared to be the coder on the golf app and y'all can be in the chat together. Now you really are vibing with friends. There you go.

Awesome, Chris. Anything else that you expected us to ask you that we haven't asked you. There you go. Awesome, Chris. Anything else that you expected us to ask you that we haven't asked you or that you can't wait to tell us about before we let you go? Y'all did a great job. I just can't wait to see what you vibe up. I'm gonna vibe some stuff.

I'm gonna get my kids vibing. This is gonna be fun. Vibes.diy if we haven't said it enough times on the pod. That's your URL. Vibes.diy. There you go. Tell your friends. Get your kids, get your nieces and nephews if you're of age similar to ours out there

and do it yourself. I mean, I'm going to throw a few things at this. I think the image gen, like the caricature stuff. Do you have a gallery of where like, cause sometimes one idea just strikes. You gotta have something to start with and you're like, Oh, I can just change that. Like the remix thing. Do you have a gallery out there? We can just change that, like the remix thing. Do you have a gallery out there?

We moderate the stuff that shows up on the homepage. So we essentially have a carousel. Every time you hit refresh on the homepage, you'll get some new choices. There's some ones that illustrate how you do real things. If you have in your mind, as many of us, you know, dorks have a mental list of APIs that have cores and don't require an API key. I've got, you know, for instance, a vibe on there that is using Algolea's hacker

news search to just bring you the vibe code news. So go remix that any API that's got that kind of access control, super easy to use. And then, as I said, with this backend support coming soon, you'll be able to drop your own API keys in. And what's the best way to get in touch with the creators of this, yourself, your team, your community, is there ways that you can say, hey, I love support for this, that, or the other thing

and you guys don't have it yet, feedback or a Discord or a Zulip or anything? It's in the Discord and the Discord's linked from our About page. All right, go forth and Vibecode with friends, especially once that multiplayer feature gets toggled on. Let us know, Chris, when that's out there, because we'll definitely give it another shot.

Probably another shout out as Adam and I could Vibecode together. You know, I could be I could be changing his design as he's trying to add features. That'd be fun. I'm looking forward to the day when your guests are on here with things they made in Vibe DIY. That'll be amazing. That will be full circle, won't it? That's what I'm trying to talk about. That's cool stuff.

That's the way it's got to be done. I can't wait. Thanks, Adam. Thanks, Sharon. Super excited to be here. And yeah, remember, everything we do is open source, so you're invited. It was a fun journey into Chris's past and really cool to see how everything connects.

You know, it's really wild how someone like Chris Anderson, someone like him who can lead the way with Couchbase, CouchDB, all the NoSQL movement and the just what went on there. It was just a wild time to be a developer. It's always a wild time to be a developer, honestly. Fireproof is open source. You could check that out. And of course, vibes.diy, get it on, do what you gotta do.

Make something cool, share it with somebody. Vibe it up, baby, vibes.diy. Okay, friends, we're in Denver. changelog.com slash live. Buy your ticket now. July is when it happens. Next month, you got a little bit of time, but do it now.

Don't wait, don't procrastinate. Super inexpensive ticket, 15 bucks. Come see Change Log IRL, a live show, a live launch, and all your fun favorite people will be there, me, Jared, Gerhard, Jason, Brake Master, and some mystery guests. It's gonna be fun. Meetup, live show, then a hike.

It's the best. Again, changelaw.com slash live. Of course, a big thank you to our friends over at Depot, depo.dev, make your builds faster. That makes you faster. It is so cool. And of course to our friends over at Retool, retool.com slash changelog.

And to our friends over at fly.io. We're big fans of Fly, you should check it out. If you're not on Fly, you are missing out. And of course the Beat Freak in Resident's Breakmaster Cylinder, bringing those beats every single week. Well, that's it. This show's done. We'll see you. Thanks for watching!