PlanetScale's Postgres Launch with Sam Lambert

We're probably going to do $1 million of ARR for the Postgres product by the end of next week, and we haven't even opened it to anybody.

How's it going?

Oh, I, I've never been so busy in my entire life.

You were just traveling, right? You took a little break.

That was just stupidity, just going and doing— like, you know, my wife's from India. Her parents are doing their 40th anniversary, so like, we don't love going to Delhi every time, just like smog and, you know, you're just stuck inside or whatever. And so we met in Italy, but her dad like booked like 3 different locations in Italy, and we haven't ate. We haven't ate. Do you have a baby now?

No, uh, one more month. Oh, congratulations.

It's going to ruin your life, but eventually, eventually, it's like whatever that weird illness cats give you, like you eventually kind of like them in the end.

Yeah.

But if you need any like help or advice with that stuff, ping me, by the way. I've like—

Yeah, will do.

Been fully through this journey. Anyway, so like with a 10-month-old, a heat wave in Italy, and then British— we had 8 flights for the whole trip. British Airways canceled 3, delayed 3, baby ended up hospitalized. We pulled the launch of Postgres forward a week. Everything just went fucking been crazy and basically got back super late on Monday. And I just have slept like 5 hours since Monday. And it's all been good. I mean, it's just all been— everything— I'm happier than I've ever been. It's been just unreal. We're probably going to do $1 million of ARR for the Postgres product by the end of next week, and we haven't even opened it to anybody.

Wow, that's really crazy. Yeah, I mean, 8 flights is crazy. I mean, with how flying is these days, We had a bunch of travel we did, and then we hit like 6 months into the pregnancy, and I'm like, this is awesome, 'cause now we can just say no to everything and no one expects us to travel, probably for like at least a couple years. Uh, yeah, I'm just like terrified of flying now.

Has she mastered the poking the belly out to get through lines and just get pro footage? 'Cause that's like a really good one.

Yeah. I mean, she's, she's like due in 1 month, so it's pretty obvious at this point. Yeah, people will treat her real nice, which is, which is good. Me by proxy too. So you launched the Postgres. I think the Postgres launch was— I remember you were literally talking about traveling. I think you were literally flying that day or something, right?

Yeah. Yeah. We, yeah, I was like, I landed at Heathrow, got into the lounge like 3 minutes before the call internally to launch happened. And so then like we launched it. I was just dying from it. Like, I just— it was— I was out of it completely. And then like, um, I just got drunk in the lounge because like the immediate reaction was just incredible. My phone was just like, um, lighting up. Twitter was going crazy and just knew immediately that people were really stoked. And so I was just necking Pimm's Cups.

Yeah, launch days are crazy because you just get so overstimulated. You're getting hit from like every angle, from Twitter DMs, texting you. So it's So it's always a lot of fun, but like by the end of the day, I'm like, I'm like so overstimulated that I just need to like cool it and like, and force myself to go to sleep.

Yes. Your dopamine, your dopamine receptors get like all jacked up. It's kind of crazy.

Yeah. So, uh, I guess, yeah, tell me a little bit about the history for this. Cause I think a lot of people were pretty surprised cause they know PlanetScale as the MySQL company and this dropped out of nowhere. and surprise everyone.

So, yeah, I mean, never say never, I guess, right? I mean, it's the thing. Uh, ever since the Metal launch, we've had like just some crazy, really crazy things happen. Um, it's a mix of just tons of demand. So like on the day of the launch, we just, you know, loads of people came in. I mean, it's accelerated the kind of the test business even more doing Metal, which is great. And we've moved some giant— like, the largest Aurora Limitless workload in the world now runs on Vitesse on PlanetScale, and they migrated to us. That was really cool. But alongside some of that demand, in fact, by about 10x, we got loads of Postgres people coming in. Like, the CEO of a very well-known company with way over $100 million ARR came to the office the next day and just sat down and was like, you have to do that. Like, here's how much we spend on Postgres. We're the only— you're the only people we would migrate self-hosted Postgres to. And I was just like, okay, this is nuts. Like, but basically on the day we had the metal party, the office was like full of hundreds of people. People were really excited. And after all that demand and just seeing the whole day of like smaller AI companies just like aping onto the platform as fast as possible, I kind of remembered what it was like to be in like an early startup where things are just— everything's good. Like, we did this at GitHub, like we moved to Fastly like in no time. Yeah, it's great. We would just do like insane shit because it just didn't matter. Um, and I just watched that happen. And so at the party, I went over to Nick, our CTO, and I just said, you know, we have to do Postgres. He went— and he'd seen it all day too, and he just was like, yeah, I mean, it can't be harder than anything else we've done. Um, and so we knew that night that it was happening. And then pretty much the day after, we told the whole company like, you have 2 weeks to wrap up any of the, like, overhanging stuff we needed for the Metal launch, to do a couple of, like, just some extra import stuff that we had been finishing off anyway. And then, like, we're gonna have a hack week, um, where we're just gonna kind of try and see how far we get. And because, like, when you just, like, look at these problems, they look really insurmountable and difficult to do things. You're just like, oh, this could— it's just completely unknown how long it would take to actually go and get this done. And so we had this very short all-hands at the beginning of the week, which was just suspend belief, we may never to do this, like, who knows? Um, but let's see if we can get Postgres into production, in our production, and being able to connect to it. That was done by Tuesday afternoon.

Nice.

And we were like, oh, okay, well, that was good. And by Friday, we had high availability and failovers done. And, um, the demos lasted 3 hours. Like, we did— like, we asked everyone, like, to demo what they'd been up to. The marketing team made fun videos, like, and, like, all of these things, like, Even like Holly just showed people how many inbound we've had over the years asking for Postgres, just like little things, everything like Postgres related basically.

Yeah.

And everyone just did, like we got insights working really nicely by the end of the week and stuff. And the main feedback from everyone was like, I can't believe how fast we got all of this done. We can do this. And we just said, right, we're shipping. We're going to be in production with real customers live on this by July 1st. And we were ahead of schedule even then and got it done. And so now there's real actual Revenue generating companies with high workloads running on the platform and are just really happy. So it was very fun.

Yeah. So okay, I didn't realize it was that quick. So because Metal launched, was that like within the past six months? Like when was when was that launched?

May. I think it was May, wasn't it?

Okay. Wow.

That was pretty quick. I'm so like dyslexic for dates. I just forget everything. It's like in the past, like it could have been any time in the yeah.

Yeah. I mean, especially when you're working at startups, like time is just like a weird concept. You both can't believe how long it's been and how recent certain things were. Mm hmm. Mm hmm.

We actually have a good amount of Postgres experience in-house, actually. We have like ex-Century people, ex-Gocardless people who all have run really big Postgres workloads. So it's felt like pretty natural to go and do this. And it's been really fun, you know, to work with Postgres more. Um, it's just been great.

Yeah. So I guess talk to me a little bit about— so there's obviously stuff that, uh, was probably unique about Postgres, but I'm curious about what was shared because I'm assuming you guys have some kind of operator that's just good at managing database nodes generally. So how much of that was able to be reused? Like, what could be reused? What was— what couldn't?

A lot. And then all of the infrastructure that surrounds that operator, right? So the post we released about extreme resilience, did you see that?

Yeah, I did. I didn't read it yet, but I did come across it.

It's basically just a list of rules that we follow for how we do this. All of that stuff underpins the Postgres product too. So the operator is incredibly mature. The way we run many, many different Kubernetes clusters that don't share any backplane, like all of these kind of layers of very resilient infrastructure are kind of done and work really well. Our edge, like, you do remember those, there's those benchmarks out there that essentially just like a latency test for getting to the database that compares us to like Neon, Supabase, and everyone. And we were just like so much faster than everyone else because our edge infrastructure is exceptionally good because we have people that have managed the edge networks of top 100 websites, right? So it's all of these things that affect the database that we've done really, really well. All of those things get pulled in. Our backup system that can back up databases at line rate, restore backups at line rate, all of these things that we've done for the hosted Vitesse product were there. And then of course, the operator being incredibly mature, having done hundreds and hundreds of millions of failovers. And, you know, it's all that. And, and so we, we— the other big design decision we made was to not do what we did previously, which was do like MySQL hosting using Vitess. Um, and then you could like scale up massively on Vitess, like our very large customers have done. We've done 2 products this time. We've done what we internally called Horizon. It was called Horizon so that we could ship JavaScript in the UI without people knowing. It was Postgres, so we could walk around San Francisco without talking about things without people knowing PlanetScale was talking about Postgres.

That's funny.

Yeah, right. Like, it was a very secretive product. We NDA'd. So there's about 50 people, including you, under NDA before this announced, just so that we could talk to people about it. And we did genuinely talk to so many customers of other platforms to kind of figure out their pain points and their problems. And yeah, so there's that. There's the unsharded, extremely fast vanilla Postgres that runs on Metal. And that means that pretty much if you use Postgres anywhere else, you can just drop it in and it works. We're still finishing up kind of getting some extensions installed in a way that won't get people hacked, you know, like the kind of approved list that Amazon has and whatever.

Yeah.

And yeah, and then we've got rapid development on the sharded solution, which you will be able to pull in when you need to shard, and you will have the trade-offs that sharding brings. Uh, we made people live with Vitesse's trade-offs when they didn't need sharding, and that was a mistake, and we're undoing that mistake for this product. So it's actually 2 products that we've been building.

Oh, I see. Interesting. And then if you do need to go from one to the other, it's like, it's the same as coming in from like an external database, I'm assuming?

We'll make it a very seamless online process, the same way like in our UI right now with Vitesse, you can reshard, you can You can migrate to different separate keyspaces. You can do all sorts of those things. We will make it extremely easy. So like right now, there's a bunch of folks migrating over that have self-sharded Postgres inside their app layer.

Oh, interesting.

And they're just swapping— they're swapping us in as those shards, right, to replace Aurora. And then we will then work with them to make sure that Nova meets their requirements, um, and we'll work with them to ramp that up in production. So really, like, if anyone's interested in sharded Postgres, the fastest way to get to that is to migrate to the current platform. That means our Insights product can see your queries, do all that stuff, and then we can manage migrating you to a sharded solution very, very quickly after that.

Yeah, that makes sense. So in terms of the sharding stuff, this is like a from-scratch new project?

Yeah, yeah, it had to be. It's like all of— we are the Vitess maintainers, We run Vitess at significantly higher scale than anybody else, with the most amount of customers. And so Vitess's improvements and learnings are what will feed the Postgres version, but it doesn't need to be Vitess. That would be like a bad architectural decision to go and do that. And it would violate our fundamental kind of architectural principles that we have.

Yeah, but I guess conceptually is the approach that Vitess takes, like architecturally, it's effectively going to be the same. Same setup for Postgres, or are there any Postgres quirks that make it different?

No, there's some ways that it'll be better, some ways it will be less optimal. Postgres has some really nice elements to it that make, you know, using its parser, using various features really nice. Its replication story is nowhere near as good as MySQL's though, so we'll have to do some work there to kind of paper over that. It's just gonna be different, but architecturally, yes, pretty much the same. And to achieve all those same goals. And then it's really gonna be that paired with the operator that gives you the overall result. And that's the— and all we really care about at PlanetScale is the result, which is extreme scalability at high, true high availability.

Yeah, makes sense. So I guess the other piece of this is the Postgres connection pooling, which has always been a massive pain in the ass.

Yes.

I'm assuming the— you guys already have like a connection pooling layer? And you just adapted that to Postgres?

Yeah, we actually have now, we have a new proxy called ps_bouncer, which does the— we called it that because everyone's familiar with PgBouncer, and we just want people to know that's what it's like there to do. Um, it is faster than PgBouncer, it's, uh, easier to make more highly available, and it will do like buffering, uh, so that you can do actually like online migrations, unlike switchovers and failovers in the way that we do with Vitesse. So I don't know if people know that we kill nodes underneath our customers' traffic constantly. It's just happening while we chat. We have to roll through millions of them throughout a month. Every single node at PlanetScale dies after 29 days. No node ever gets to 29 days of life. It either gets hit by a maintenance window, it either gets replaced due to an underlying failure, or we need to do a version upgrade or something, the answer to every single one of those things is a PRS or an ERS, which is just killing the nodes in that cluster and having them reconverge, thanks to the operator. To do that, we need to be able to stall queries in flight and redirect them to the right place. We have achieved all of that for Postgres already. And so that proxy is very good. It's very low latency. It's lower latency than any one we saw at any, at Amazon, Google, any other provider. We can get you talking, we can get you from your frontend to the database server with the least latency of anybody. And then it can also do all those special things for you. It can route you to the right replicas in the right AZ and do all of those magic things. And that itself is now undergoing a partial rewrite to make it even faster and have an iterative path towards sharding.

Yeah, cool. That makes sense. Yeah, I think the The connection pooling stuff, I mean, even in AWS, like if you, which is, you know, one of the most used Postgres services, you can't just spin up Postgres and be done with it. You got to also spin up their, their data proxy thing. And I'm just like, it's been so many years of this. Like, how is this not just integrated and solved on their end? It's like, you, it's like not, it's not optional. Like you need it for sure.

Postgres is going to have a long journey. The Neon team actually started this and doing the good work to go and grind this out, which is moving to a thread model for Postgres.

Yeah, I have seen those threads.

Yeah. And I, like, shout out to them for doing that, right? Like, it's a— it seems like the third rail. I think a lot of people are very unhappy with that idea. But it does seem to be— it is something that needs to change. And, you know, we've seen, you know, we've had to tune, like, memory management for Postgres as well. It hands a lot off to the operating system in a way that MySQL doesn't. So then the operating system isn't aware of this, always managing a database server, right? So, like, You just, it's just different in terms of like things we have to do. Luckily we have people that have worked on the kernel, so you can just, we can figure those things out too.

Yeah. Nice. Yeah. That Postgres thing, it's such a longstanding issue. I remember there was like a thread finally talking about it, but obviously when you have a codebase that's like as old as it is, that was effectively single-threaded, you just have globals and stuff everywhere. So that just seems like a complete nightmare to work through. It has to happen and it will get done, but Yeah, painful task.

Yeah, it's manageable. The replication story is worse and harder to deal with. But again, I'm sure that will get better over time. We'll try and help there and we'll compensate in other ways.

Yeah. So I guess let's talk a little bit about like the kind of the business of Postgres. So for me, you know, whenever I've used Postgres, it's always been Aurora and kind of nothing else. You guys seem positioned, and you've been positioned for a while, as like the better alternative to Aurora. Is that kind of how you're thinking about the Postgres side as well?

You know, I looked at everyone other than Aurora and it just doesn't matter. You know what I mean? It's like very easy. It's like hitting a child. But Aurora is like a real product, you know?

Yeah. Do you have any numbers on how big they are, like business-wise?

Yeah.

Well, I don't know how much you can share, but Yeah, it's a major part of Amazon's business, is that.

And Amazon as a whole company, Aurora stands out as a major contribution to their entire business. And it's a very good product. Like, it truly is very good. We are provably better, and that's been proven through our customers' stories. Like, if you go and read the case studies, very large Aurora customers have migrated to us. seen cost savings, seen significant performance improvements, and most importantly, reliability improvements, which is always stunning to me. I spoke to someone very senior, formerly from AWS, last night, actually. And he was like, yeah, they know that. And it's kind of very stunning, actually, that that's the case, that they've achieved that. And to me, I kind of pinch myself. And then you're reminded that this is the technology arc through and through, right? This is small companies with advantages can eventually win against very large ones. Um, and so it's been really interesting. And actually, the amount of email— like, you know, we, we've been doing this kind of ramp-up from like moving people from the smaller providers over to us, and that's been like really easy, and they see an immediate benefit. And we're just going through building our confidence, building everyone else's confidence. I've been surprised at how many Aurora people have reached out, like really wanting to dig in and try it. Um, so. It's very cool. Really cool.

Yeah, it's weird given how much attention Postgres has gotten. Uh, you know, Aurora has still been the default for so long. So it's interesting to see that kind of finally start to change. The other thing I was going to ask was, uh, obviously in this initial phase, I'm assuming it's smaller companies. Uh, I'm sure you have bigger companies like looking at, looking to move to, or maybe have already moved. What do you think about this whole, there's like all this talk about how there's some link between Postgres and AI companies. Uh, there's a lot of like framing in that way. Like, do you have any thoughts on, on that?

The majority of AI companies are getting founded using Postgres.

But like, is that unique to AI companies, or is that just startups in general?

It's also— I think it's just startups in general. I mean, part of the thing that was very revealing to me was talking to like Aurora users or whoever and just saying, why did you choose Postgres? And it's just like, default.

Yeah, exactly. It's a—

yeah, everyone else uses it. It's like, yeah, I mean, that's it really. Like, no one— most people's initial choice for a database is just ask a friend, use whatever the tutorial framework or thing you're using is, right? Um, when I realized that was the criteria, I was like, okay, then if we have something unique to offer this market, we should, because there's just so many people building there. Um, we had to obviously convince ourselves we had something unique, but Yeah, I think it's just the default for startups. The thing is, we've seen from AI companies, and I think we'll probably talk more publicly alongside them soon, that one of the largest AI companies has migrated to Vitesse now on planet scale with a massive workload. Because the AI companies that are taking off, it doesn't matter to them what— there's a religion just goes out the window. They're in a— They're death wrestling with ogres, to use the Succession quote. But they're all fighting each other on models and GPUs and trying to get GPU capacity and trying to do all of these things. Why solve database problems alongside that when it's an undifferentiated problem that another company has fixed? So at the bottom end, yes, lots of AI companies are migrating and wanting to move to our Postgres product because they just use it anyway. And then at the upper end, because we are this database that's been powering very large consumer products, for a long time, um, they now are consumer products, right? They have to just be online, up. And, you know, one of the AI companies that's migrated to us has— is powering just a ton of like real just consumer usage that's nothing really to do with the model or whatever, but it's just like we have a ton of users now.

Yeah.

And we create a lot of writes. The thing that's very different about the AI companies is their write volume is significantly higher and of different proportions than traditional OLTP workloads because they're doing chats. Yeah, they're storing like, they're storing what the user asks for, what the model talks about, what comes back, like all of— you just want to keep everything. And in a relational database that can take a lot of writes is a really good place to store, to do that. So yeah.

Yeah, that's true. I never thought about it. I guess it's inverted. I'm assuming people don't read. It's like each user's generating a ton of data and they're probably not reading. it all the time, the invert of the normal workload.

Yeah. Which if you compare that to big consumer OLTP workloads like Twitter, for example, one tweet is like a single insert that fans out to millions of reads potentially. GitHub's the same. You read or generate a read on GitHub way more than you do actually write something. And so most OLTP workloads are 100 to 1, maybe even 1,000 to 1. That's the complete opposite for certain AI companies.

Yeah. I mean, this reminds me of another thing I think about a lot, which is if I kind of just step back and look at how easy this stuff has gotten as an end developer. I remember back in the day, we had to like, we like ran up our own Cassandra clusters and maintain that ourselves. And it's hilarious how like, that's not something I would even ever probably in the rest of my life even come close to thinking, like forget even There's a side of maintaining my own, you know, self-hosting this stuff. There's also like even going to those technologies, uh, like that, even that has, has gone away where, you know, stuff like PlanetScale can kind of just handle a pretty wide range of workloads without needing like specialized tools. So yeah, it's crazy how the database part of applications has just gotten really, really boring and invisible.

Which is good.

Yeah.

Like you said, it's exactly, you're right. Like database problems can just interrupt anything your company's doing. Like you can have outages, but even then, if you you wake up in the morning and last night's backup didn't complete, you, you are not— you can't do anything else that day. Like, no matter what you had planned, if you have an issue now, you could ruin your business. And so, like, those things not happening is, you know, I wish more people had to actually do it, um, and they would understand. That was actually a lot of the debates I would have with people on Postgres arguing with me. I realized they've never hosted a Postgres cluster. And I, I used to ask people, like, talk me through an online version upgrade for Postgres, and they just couldn't. They've never thought about how to do it. And you kind of realize that these are just problems. Like, one of our issues now, and it's a nice issue to have, is so many people are asking to migrate and then are not quite sure how they would do it online. They have to, like, plan a multi-step process. So we're doing a lot of kind of education to help people understand how to do that. That's going well, but it's still— it's not an easy— it's not just any of this database stuff is, like, not easy, and people find it very scary. And so they should, you know, you can Do some real damage.

Yeah. So what's your guys' goal? Like, you're trying to become like an IPO? Like, that's the— that's the path that you're headed down?

I don't know. I really don't. You know, I don't think that far ahead. I always feel like you just have to earn thinking about these things, right? You know, I just don't— success and what my definition of success is, is far off, and it's a daily grind of working really hard to get there. There's a lot of people that fantasize about that, that end and that exit. And I can tell you that—

It's more of a side effect of everything else.

Yeah, yeah, it truly is. Like, it's the people that don't focus on that, that just focus with what they love, um, end up accidentally building the best companies. Um, yeah, I just don't think about the end. The end is going to be always bittersweet.

Um, Yeah.

No matter what happens, whether you get bought, whether you IPO. And we're far off, um, the goals I have for the company, so I don't think about it. I do, but I just, you know, being a profitable company that is now exceeding all of our financial targets, um, is a very powerful position. I mean, I— we have competition all around now in the Postcodes world. I just— we just have to wait. I mean, every customer they get shortens their runway. If we never sold another dollar of software, we would still not run out of money. Um, and so that's extremely strategic. Like, it's just very— people have to now just— you see companies like throwing away tons of brand equity to just chase vibe coding. Now, there's nothing wrong with vibe coding. There's truly nothing wrong. But there's catching these tailwinds, and it feels extremely exciting. But at the end of the day, like, you've got to truly capitalize on that and inverting a heavily negative gross margin business in that world is an almost impossible feat. If we are positive gross margin business that builds for really large customers that stick around and are happy, strategically, that's something that's really powerful and will build a great company. So we just keep doing that over and over again and see what options come up, basically.

Yeah. I guess to phrase my question another way, like, is when you think about your goals or your dreams, for the company? Is it that like every single database workload is running on PlanetScale? Like, that's the, that's the hope you have?

Mm-hmm. My— I find all of the things I— that make me the happiest when it comes to, you know, things that happen at the company is when products I like or that lots of people like run on the platform.

Yeah, that's really fun.

It's extremely fun. I mean, every day I pay for coffee, I know that it's hitting a PlanetScale database. you know, every message I send, you know, it's just, it's really fun. I mean, it's just really cool. And there's like companies like WAP, right? Have you seen WAP? This is a New York company. They're doing a kind of, I guess it's like a Shopify for the world of creators. And they're just absolutely crushing it. And they've paid out billions of dollars.

Oh, yeah. I think I saw you talking about that.

Yeah. They're just a phenomenal company. And I feel so inspired when I like work with them and go and see them and work at their office. And it's just like so cool to be seeing their journey and seeing them grow. And like little moments, like, you know, kick.com runs on PlanetScale. And I remember them signing up at the beginning of them developing their app and had a small, like small tiny database and now are like a good size. And then they have a Formula 1 car and you're looking at that and you're like, yeah, you just, you just like, these are really satisfying. It's a really good feeling. You know, I want to keep going and going and going. And to me, it's also the trust, right? Someone looks at you and says, this is a real— we have a real company, we're a public company that's running on your platform and they trust you to do it. To me, that's like the best. Absolutely. It far exceeds any kind of money, monetary goals.

Sometimes when I look through our like customer list and I kind of like visit their websites and see what they're about, the other feeling I get is you get a sense of how big the world is. Like you just see all these companies doing all these things that you'd never even heard of or would think of. And there's— you imagine, oh, there's like a team of like the people in an office and they think about this stuff every day. And yeah, you just feel really connected with how large the world really is. Really good feeling.

Yes, we have a few companies that, yeah, like, I just can't believe, um, how much generate— like, how much database workloads they generate for the things they do. And they might be in other countries and in markets and products that I've never thought about for myself, and it's just really fun to kind of see that.

Yeah, yeah, because otherwise you're kind of like in a bubble, but like when you see that, you realize how big the world actually is.

I know the service at Blizzard that uses Vitesse, so when you load Call of Duty, I see that service load, and I know it's loading. And yeah, it's just very like— that's just very cool. It'll never not be cool to me. And like going and visiting their headquarters and seeing Yeah, I was a WoW player for forever, and you're just like, okay, well, now I'm visiting Blizzard, an actual customer of the company that I'm running. It's just, you get these pinch-me moments that are amazing.

So, I was gonna go back a little bit. You were talking about the vibe coding thing. So, obviously, there is a trend, I'll call it, right now, where there is a bunch of companies building platforms to help people build apps from 0 to 1, and, you know, deploy them, and that obviously includes a database. And it's kind of a rush to like try to be the company that powers these things. Yeah, I'd love to hear a little bit more about how you think about that. So I'll say, I'll tell you my perspective first, which is I don't know whether this trend is valuable or not, so I'm gonna ignore that part of it. Let's take for granted that, you know, this problem needs to be solved and it's worth solving. Uh, the approach right now has been, at least it seems like, uh, has been to spin up real instances of infrastructure for each one of these services. To me, that feels fundamentally the wrong approach because this is more like millions of apps that get very little traffic each. And that model just doesn't seem to make sense. But yeah, I am curious to hear your thoughts.

So before I answer, I'll also do the same caveat, which is I love AI. We have very large users of AI and very serious users of AI tools, and we have very serious AI customers. And I I code using AI.

Yeah, it's fun.

Constantly. It's amazing. I mean, I love— there's— I love being in the era. I saw someone use the term of like plastic software. I think it's really cool. Like an app that you only need for a month, right? So you just like make it really quickly. And like, you know, I've got a few of them on the go right now, like a dashboard that I wanted to build, right? That just connects into our metrics cluster, pulls a load of things. You get like something interactive and fun, right? You're just like vibe coding that stuff up. Yeah, it's really good. The stuff with Cursor is incredible in Slack where you can just go, fix this please, and it does it. It's just like mind-blowing to me. It's like real actual like magical bicycle of the mind level, kind of like these are computer moments that you'll remember for the rest of your life. Um, so I think that's sort of cool. Uh, Convex has a vibe coding product with Chef, and that is doing like incredible things. And the primitives they built into our product make that stuff really easy and making those situated PlatStack apps really easy to use. I do that, but I do, however, in answering your question about like whether these— the infrastructure is right, I do think there's like a lot of usage that's just going to go away eventually, right? Like there's, there's a max. Like we've had things like Squarespace, Retool, all of these kind of tools that have allowed you to simplify automation or just simplify website building and all those things. And then you just realize that people don't have that many problems or ideas. Like, automation has in no way been deployed at the level it could have been. Most people don't want to. Most people aren't creative. Like, you put people in front of ChatGPT and they don't know what to ask. It's the same with Google. You ask someone to Google, like, who's been an internet user for years, they're just very naive, kind of. We've had our fingertips on this stuff. So I just, you know, these companies that are growing very quickly, they're also churning through customers very quickly, right? So they go through people trying it out for a month, and the next cohort comes in, and You know, it's people just like fuck around and whatever. They are using databases to do this, which obviously you would do. The thing I don't understand, and I was in this position when I saw a lot of like dedicated vector databases or, you know, other technologies, it doesn't quite gel for me why it wouldn't just be SQLite that you would use, right? And people don't understand the difference between a database server and a database engine and what they do. Database servers contain an engine. but they do significantly more things with the assumption of a, like, a multi-tenancy case, so more than one client, right? And, and the work— for those that don't understand how these things work, I will tell you that it's probably often 2 to 3 times more code problems, bugs are associated with the server part than the engine part, right? In fact, the engine part can be much, much simpler.

Yeah.

And it's because of things like MVCC, multi-version concurrency control, like Making sure that when you're writing to the database, it doesn't clash with me writing to the database, making sure that both our connections are terminated to that database, that authentication happens, and all of those things. It's all just to assume more than one user. These AI apps don't need that. So all of the things that come with a server, even if that server can scale to zero, it is still a useless overhead. And I don't understand why you couldn't just embed SQLite databases everywhere for doing this like we already have. And I think a lot of people are probably coming to that conclusion too. It seems to be the common take, but who knows? I've been wrong about loads of things. I, you know, I also have biases, so I mean, it could just be completely wrong.

It's also funny because people have been bringing up SQLite forever, and this is like the one time where, hey, like, it actually is a good thing to bring up, but nobody seems to be really exploring that path.

Yeah, there's 2 things that obfuscate anything in tech, which is if you're giving things away for free, that's just impossible to even judge then. You know, I still get— and it's picked up even more in Postgres— but I still get 10+ emails a day. Why don't you have a free tier? All these sorts of things. And it really shows you— no, but it's like, if you could imagine every single one of them would be on your product and would still would never pay you money, it would change a lot of the metrics that you could share, right? Like, it would—

Right, right.

Yeah. When you're offering free computers or basically giving $3 away for $1 on the internet, you're going to just attract a ton of usage, which is fine. So that's one thing I think maybe that because there's people giving away those hosted databases for free, essentially they're using that instead of SQLite or whatever, and because it's already built into the product. Yeah. And then just the hype cycle of their marketing budgets paired with that, you just can't really tell.

Um, yeah.

Yeah. I mean, SQLite on the server I think is a bad idea, uh, but embedded it's a phenomenal database. I've used it. I use it I've got so many SQLite databases because for that usage at all, I'm not trying to build a server for it. And then everything else is really simple. I think it'll win out in that way, but it will not change that. SQLite winning in that world will not change the balance sheet of any particular company. SQLite is an incredible, um, open source project that is good as it is and is perfect for this use case now, you know.

Yeah, the, the giving away money is a funny thing because The way I see it is there are cases where it works really well, right? Like subsidizing something. In all of those cases, there's some kind of network effect involved. So if you like take the classic example of Uber, yeah, they subsidize the hell out of that stuff, but the end result that they get to walk away with is a gigantic network of drivers and all that stuff.

Yeah.

Uh, so like there's— it's basically once that's bootstrapped, it's impossible to really compete with them. I think people take, see that and they kind of apply it to everything where subsidizing a large user base for a dev tool product, once you stop doing that, they can just leave because there's no, like, there's nothing tangible. There's no network thing that you've built up. Um, and to me, this is like, so it's like really obvious and basic, but it just keeps happening. And it's happening now with the AI stuff as well, because you're starting to see I think it's a very old lesson, which is if you're a business and you have a cost and you pay a certain way and you're charging your customers in a way that doesn't match that, that's just not going to work. Eventually it's going to be something that's not sustainable. And you're seeing that with AI companies now where they give some kind of fixed price monthly thing, but then their costs are usage-based and they're starting to have to deal with that. And again, these are like productivity products. They're not like networks, social networks, or like marketplace-type businesses. Um, but yeah, like this now, we've seen so many iterations, or I've seen so many iterations of this now. Keeps happening.

That was the thing that was very surprising when we were looking at other database companies, that I was learning from their customers that their bill wasn't based on usage. It was built based on how many projects they'd set up or whatever. And I was like, that's a very— that's the— a very unusual profile for database companies to have in terms of revenue. Ours is purely, purely usage-based. Like, there's, there's people paying us north of $1 million a year, and they have 3, 3 accounts. Like, they just have 3 administrators for that database that just access it.

Right.

So it's completely inverse to other database companies that are kind of caught up in this AI hype, which is like, extreme user growth and then no monetization of, of any particular growth. And that, that, that— I mean, the whole world could shift, but there's a really good— it's like a Substack post called— I think it's like Enterprise Businesses Are Fat-Tailed. People don't understand that like there's this extreme power law to everything in tech, especially revenue. And if you're trying to get to $500 million of revenue, you need some huge building blocks to do that. You can't just sell $29 projects, um, yeah, that have a database as part of it. That doesn't work. You need, you need certain people paying you a lot of money for a lot of consumption that they cannot move from. They, like, they, like, it would have, you know, the reason people migrate when people change free tiers and stuff is because they don't have problems. They don't— if they're not, like, if it's not going to interrupt a paying customer of theirs, it doesn't matter, you know. This is really interesting. That's kind of another interesting thing we've seen from this Postgres launch is I've got so many people and I've been giving them access that they're on other providers that haven't launched their product yet. So they're like, please let us in before we launch next week so we can move. And I'm doing that because it's the second someone, like someone depends on your service, that move is orders of magnitude harder.

Yeah.

So I'm just letting those people have access so they cannot have that problem in the future. But it's just been very interesting to understand a bit more closely how certain companies are being built. Um, it's very interesting.

Yeah, I mean, the rules are actually really simple. It's just build a product where no one has to eject from. Your customers shouldn't have to eject from it, and then charge them in a way that matches your costs. And if you do both of those things, you're fine. But yes, weirdly, it's rare.

Yeah, I mean, the retention numbers of databases that don't let you down are very high, it turns out, you know, like very, very high. You people just don't take the risk after that's, that's happening, which is really nice.

I mean, that was another thing that I was, uh, reflecting on when you guys launched the Postgres stuff, just The response is so funny. It's just like, oh yeah, like of course I'm gonna use this, or like finally, or like everyone's really excited. And you don't really have to— it's not like you have to go out there and be like, use us, we're very reliable. Like, you know, we— you don't have to really say any of that. It's already understood that, you know, if you guys do something, there's gonna be certain characteristics that are gonna come with that, and that's kind of assumed and And everyone trusts that's going to happen. And it's such a good position to be in brand-wise. You guys have an actual brand. That's what having a brand means. People see your name on it and they know certain things about your product.

That was the craziest thing to me. And I said a huge thank you to our marketing team for achieving this, which is people really got it. They heard the 2 words next to each other and they knew what it meant. And I was really surprised. I truly was like, wow, really, we are synonymous with these qualities and are the only qualities I care about for a database. Um, and I was like really proud of that. Like, it was just awesome, actually. Um, yeah, that was real weird, actually, to see, to see that. And I also thought we— I would have spent the week with people going, uh, you said you don't like Postgres. Like, no, you're a hater. One, that's just not true. And 2, um, no one's really saying it. They're like, oh cool, like you're gonna come and do this thing in the Postgres world, there's a space for that. So we're really happy to see it. So that was nice. And, um, yeah, we've, we've done a lot of hiring since that launch. There's people from Postgres companies that are like notable people in the Postgres world are joining the company, which is nice to see too. And it's really cool.

Yeah, yeah, I think that's a good lesson that people should take away because it's not like you guys built that brand overnight. It's not like you did a few things or spent money in a few places that achieved that. Like, that's a goal of every company. Like, not every company has a brand, but if you do have a brand, it's like a ridiculous advantage. Uh, and it's built slowly every single day over years, and it doesn't really look like anything until all of a sudden it does in moments like these. Uh, and I think companies feel like— seem to think it's like, if I spend the right amount of money with the right design agency and I like have the right name, and it's, it's really There's short-term things that you can do that look like you're building a brand, but it's all just sugar. It's just this slow, long-term thing. And like everything else with startups, but this one especially so.

Yeah. And it takes a very long time in databases. You just have to grind through winning trust with people over and over again. Like, they're already on Amazon, even if they're unhappy. Like, I don't get fired for trusting Amazon. I mean, you know, that's, that's done. And yeah, so yeah, it just takes forever. And I think people really try and get quick hits and ruin really good brands. Like, a lot of the companies are getting trolled right now for their lack of reliability. Like, they could have avoided that by just not saying they were hyper-scalable or infinite or bottom— you know, all these words that like don't really— like, if you say them and someone finds out very quickly they're not true and then shares that that's not true, it's really negative for your brand, like really negative. And so you have to not do that. And the reason we share so much about our customers is because it's like it gives you that shine of, you know, their brand trusts your brand. So that is additive to everyone, which is really nice.

Yeah, a lot of this stuff always ends up feeling to me like it's this slow process of continuously not doing the wrong thing. every single day versus explicitly doing stuff right. It's more just like, there's all these wrong things that are easy to do. And if you don't do them long enough, then it works out. Yeah, I just feel like there's traps everywhere and you kind of over time understand how to avoid them.

Yeah, you're right. And I was thinking actually, could we have done Postgres sooner? And I'm certain the answer is still no. Like, I'm truly certain the answer is still no because it would have been a distraction. We hadn't earned the right to go and do that yet. All of those things, like, it just— infrastructure tech companies take way longer than people realize. Like, way, way, way longer than people realize. In fact, the ones that grow rapidly at the beginning aren't really, like, often selling infrastructure, you know?

Yeah.

And so, yeah, you built like a Snowflake. I mean, if you look at it, took forever to build that. And then it goes off like a rocket, right? But until that day, it's really— I mean, even look at Figma, right? They had lots of small flat rounds at the beginning of their company because they were just figuring out how to do efficient CRDTs in the browser, right?

Yeah.

Until that technical hurdle was cleared, the downstream product was not going to be anywhere. And then it goes off. And until you've done that really hard, deliberate work, then it's just not going to— It's a technical operation. That's why venture capital can be extremely positive for deeply technical businesses like ours. Like, you know, yeah, couldn't be profitable out the gate when you're trying to build something that's deeply technical, but then when you get to that point, it's very, very beneficial.

I think I try to tell people is like, when I see people starting new companies and stuff, I always ask them like, are you really— this is something you want to do for like at least 10 years? Because it's not going to be fast. There's stories of things happening fast, but For you, it's probably going to be 10 years minimum. And is this really— are you excited to give up that amount of your life, focus on this thing? Because that is what it takes.

I have recurring nightmare that like I do investing, you know, now and then, just like cut people a check now and then. And, you know, some of them have paid off. I have this nightmare that one pays off extremely well and then I, like, don't want to do stuff in the future. Like, it makes me somehow lazy. I'll tell you this, which is I worked at GitHub, we sold to Microsoft. That was financially lucrative for a lot of people, and it came with, like, extreme sadness for me. I think it was one of the greatest companies.

Yeah.

Of all time. And, um, you know, a lot of my friends I made there, and, like, you know, some of the people I'm closest to in the world that I see every single week and go for drinks and we hang out, you know. And you look back and it's like everything we whinged about, every problem we complained about was just ridiculous. We were 27 years old at one of the greatest companies of all time, the world of software development. And I regret every second I didn't just breathe all of that in and just enjoy it because that's your life. Like, it's, it's what you get to do. Um, I, I fear it ending. When people like joke about an exit or whatever, I get really snappy about it. I get really snappy because I'm like, yeah, okay, you'll be like rich or whatever. And first, everyone is already rich. And so you're just like— but then it's over. And then it's like, what do you do next? You just, you know, buy watches or whatever, you know?

No, seriously, you gotta start over. Yeah, it's, uh—

Everyone who knows me truly knows I will be at work when I die. Like, that is just how it is. Do you know what I mean? Like, I want— I love working. I work every day of my life. And I just— it's the best. I just— I like, I like doing that stuff. It happens to be symbiotic with making money. But, you know, I know so many rich people that have done nothing in their lives. I don't respect them. I don't want their opinions. I don't care about them. And they'll— they have tons of money. I realized long ago that there's no correlation between wealth, skill, or intelligence. And so now I get to work with people whose skill and intelligence I respect. I want to do that for as long as possible, basically.

Yeah, I really resonate with how you describe it as bittersweet or sad because I remember early in my career, obviously, an acquisition, that's the dream. Oh my God, I hope one day I get acquired. And then as you start to get more successful and those options start to show up, you realize that the day you say yes to that is a day where every dream you had of the future of what this thing could be is over.

Yep.

And that's such a sad feeling. It's like so negative that it overwhelms whatever, you know, financial portion of it is. Um, and I never— like, this, this seems, this seems crazy to me. This would seem crazy to me, but like when I was first starting, like I remember I would see like, uh, companies that blew up and got crazy acquisition offers and they said no to like several billion dollars.

I'd be like, that guy's so stupid, like what is he thinking?

But now I get it. It's like Yeah, and then what? Like, everything you wanted, that you were hoping for, things that you wanted to do, you don't get to anymore.

Yeah.

Um, and that's kind of what overrides everything.

Absolutely. Yeah, no, I mean it. And like, if you're going to take it to the logical extreme, like, why aren't you working on things that make even more money? Like, go fully evil, you know what I mean?

Just like work on—

no, but seriously, like, why are you— what, what dumb midpoint? If you're like, you're okay, so you're, you know, you do what I mean? Like, the middle of the, like If you're not making— if you just care about money, go and make extreme amounts of money doing something insanely boring. If you want to do something that's purely creative, then why even kind of intertwine money with those 2 things? I like both. I like the sport of building a business in a market that's highly technical. I have, like, I mean, I've— before I came to PlanetScale, before I've been doing this, I would get offers all the time to, like, be the VP engineering or CTO at a sure thing company, like, like, just a sure thing.

Yeah.

With, like, with an equity offer, it's like, yeah, I'll make $100 million or something from that company if they IPO at like the last valuation or whatever. Yeah. And you're just— but the product just did nothing for me. It was— it's just so boring. I just wouldn't go to sleep excited for the next day. You'd just be grinding through it, you know? And yeah, that's not what being alive is about.

I'm sure people listening to this are like, probably think we sound crazy, but it's just one of these things where when you get a taste of everything properly aligning in this way, it just— there's no going back and you kind of get addicted to it.

Yes, it's— yeah, I mean, and of course, like, you realize that when you do have money, right? So it's like, okay, well, you know, um, it does— there's a level of, like, earning that you do that kind of makes you feel very safe, and that's a completely different feeling. And then you get those— you get the option to go and want to work on things that you love, and that's really nice. But there's this kind of person that gets a lot of money and was never that passionate, and now they just have time and money, and they just do all of the standard boring rich people things.

Right.

And it just looks so boring to me. You know, it just doesn't look fun. I like to do the things I did when I was young. That was it, you know? Um, a good friend of mine is very successful. He still just codes and does video games, 'cause that's, like, what he always liked doing. Why do people suddenly like make a ton of money and start getting into just very generic rich people? I just don't understand it.

Yeah, it is funny. I always, uh, I mean, I'm the same way as you. Like, I know I'm gonna like die at work as well. But you know, I actually think, I think about this all the time because I've seen so many personal situations where people retire and their health just like deteriorates and their brain deteriorates. And then, yep, I've, I've seen it in such crazy ways where I feel like they, I think they literally died because they retired. You know?

Yeah.

Um, so yeah, it's kind of like a, it's kind of like a sickness in a way.

Yeah. People's like brains go to mush. They've seen this. One of the blue zones, I think there's a blue zone in like, is it Okinawa? Somewhere in Japan. And I believe like one of the accepted reasons that people live so long is that it's traditional for the grandparents that they take care of their grandchildren. I see. And it gives them such connection and purpose and a reason to live. that they're just like happy and, and have purpose in there, and it helps them stick around. Um, so it's kind of wild. Yeah. And I just like— I also— the thing that I find really nice now is helping younger people that are building companies. I always tell them not to listen to me and don't really try and replicate anything that I've done, um, because I don't feel like I've been successful yet, right? Like, until you build something really incredible, um, that meets what I want to build, then I don't feel that successful. But helping people kind of short-circuit some of the standard issues that you run into and the mistakes that you make. And I've made tons of mistakes. Like, if you could just— you ask everyone 5 years— if you listen to this, you know, anyone that's running a company 5 years in, what they would do differently, it would be nearly everything.

Everything.

Right? You just make mistakes. Like, there's no playbook for what you do. People like to think there are. They really think there's templates. But when you study the great companies, they're all off of— they don't run on a standard template. It's all weird, winding paths to get to where they are. But like, yeah, just like helping people through the undifferentiated things is like very, very satisfying. And it's also why people should always back second-time founders, because it's truly, truly— you just learn so much the first time.

Yeah, I mean, I feel like I've had to make every mistake like 3 times before I really— I've really understood it.

One of the hardest to avoid is just constant temptation.

Yeah, it's crazy.

So many different opportunities have come in in the last week. And it's just like, okay, we have to stick to the fundamentals of the things we do very well, not get, you know, and just— it's that— it's like that truly like a marathon run of just like a constant pace. And I actually think that's the most unfair things we can do to our competitors is that you haven't really seen a lull in shipping from PlanetScale. Like we have just—

Right.

At just like a marching drumbeat of pace. Like everyone can pull off a sprint to rush some shit out in a quarter, and then like everyone's like exhausted and dying. If you can just constantly march at a pace that just no one else can keep up with, it's really unfair to folks. And so it's that long game playing out.

Yeah. The temptation thing is funny because it kind of goes back to what I was saying about there's just traps everywhere and it's just about like avoiding them. I, you know, when I was younger, I imagined building a business was like, uh, like you're desperately looking for some opportunity and like you just can't find it and you're hoping that opportunity shows up one day. as you get a little bit more successful, it's actually the inverse problem. There's just a million awesome-looking opportunities all the time, but they're hiding the better opportunity that's like a little, a little harder to see. So yeah, it just feels like every day it's just about saying no to pretty great options, really, um, and having to focus on that. And yeah, it just takes so much experience to even recognize what those things look like.

Yep, absolutely. I can remember in the early days of GitHub, a very, very large company had an open checkbook to do like do a certain type of GitHub deployment that would have been a one-on-one for them, and it would have been a nightmare. And it was just so much money.

Yeah, yeah, exactly.

And at the time, it's just like, wow, that seems amazing. But if you'd have done that, it throws you off course. Same with us, right? Like, we get asked continually to just do a purely on-prem version of PlanetScale. We've been asked so many times with people that would pay extra, like, at certain times— like, it would never have happened now, but like, at that time would have like potentially doubled our revenue at the time that we were being asked to do it. And it gets really tempting. And I remember all of the experience I had from other companies that showed me not to do that. But it's that quick temptation, especially if you're not profitable. This is the— this is where those kind of like that kind of knowing your cash-out date, right? And, you know, for those that don't know what that means, like if you— any of your vendors, anyone is not profitable, that means the CEO and executive team and the board know the exact date where the business is finished. Like there is a date in the future that's calculable where there's no more money in the bank, just the costs will overlap what's left. That is a feeling that is very difficult and very hard, and it makes you act weird. And so when we released the first version of Metal, the test version, we were going to try and get it out in beta before re:Invent because re:Invent is a big conference, a massive conference, actually. if you've not been to re:Invent, you should, just to sort of confront how large the commercial side of it is. It's crazy.

Yeah, yeah.

Software engineers don't understand, I don't think, how large the commercial part of the industry is. It's a conference that could only happen in Vegas, and it's across 5 hotels.

And yeah, it's insane.

Probably a billion dollars of marketing spend across the whole thing. Just ridiculous. Oh, so anyway, so we were trying to rush it out for that, but instead I've, you know, I've told that story. Our customers were having problems with EBS. So we said to them, we'll migrate you first, actually. And that took an extra quarter to do that. But that meant the announcement was so much better because it was like, real people use this. It's real. It's GA today. It works. Um, you know, like, if you were unprofitable and had 18 months of runway, you wouldn't do that. You couldn't. You'd have to like get whatever crap out there to try and spend a load of marketing money to build pipeline, to desperately try and get leads and have something to show the next set of investors to try and get your next hit to keep alive and keep going. That's rough. I mean, a lot of these AI valuations and raises that you're seeing are based on the fact that the growth they've seen is costing them so much money, they have to raise to stay alive. It eats their runway. That's not great.

Yeah, it's another aspect of this that I didn't understand early on. you know, there's like being smart and like making good decisions. And but so much of it is about protecting your own ability to even think clearly. Like so many things can just take a smart person and just have them start making horrible decisions just given the situation they're in. So yeah, like we talk constantly about, you know, are we like thinking clearly? Like what's biasing us? Like we like avoid situations that have upsides, but, you know, would like impact our ability to like assess the situation properly. Um, yeah, so I'm like really paranoid about that. I'm like, am I getting involved with something that, that clouds my judgment? Um, because it doesn't matter how smart you are.

Nope, anyone can. And there's times when you're very vulnerable too. Yeah, right, like 3 bad things happening in a row to wake— like, you know, just the temptation. You're, you know, you're sat there alone at 2 in the morning, you're thinking about the seriousness of what you've done and what you do. Yeah, and it can, you know, it can be hard. That's why you need people around you that will speak the truth, keep you, keep the emperor's clothes on.

So this last thing before we hop off, uh, I just remembered that, uh, you tweeted reminiscing about React the other day.

Yeah.

Yeah, so I, I spent a bunch of time with Erlang and Elixir, and like React was obviously like the god in that world. But yeah, we had a deployment where we were able to embed React inside all of our application nodes and like it clustered and all the data querying was like effectively local. I was like, this is really awesome. But they died out, right? They kind of disappeared.

Yeah. I mean, that's the way of most database companies, right? There's a lot of really niche things. Um, but I have like, again, I have like a lot of nostalgia for some of the—

There was an era where there was so much database experimentation. That was fun. Every new startup, I would pick something different.

Yeah, I agree. And there's like really cool, like really cool stuff built by really, really talented people. It's like if you like cars, right, you can appreciate the ones that are just incredible powerhouses that, you know, everyone loves and they work and then whatever. And then there's the weird niche ones that have tried something completely crazy and you're like, that's fun and cool too. Like, you can— I love that about tech. I truly like— this is why I love Convex, because it's, it's actually the most kind of modern and dynamic database I've seen that is actually extremely real and practical, has really serious users, and they've kept everything else normal. And it's being built by a team of people that— so it's the ex-Dropbox MySQL team. They ran for like 15,000 MySQL nodes. So they know that, like, these are people I truly, like, respect. Some of the folks that work with them actually work at PlanetScale. Like, we kind of— everyone knows each other. And they've taken that underlying infrastructural knowledge and then built something incredibly dynamic and forward-facing on top. That to me is just— I love it. Like, I absolutely love, uh, that as a way of approaching the problem versus people who show up and go like, we're the end of SQL or whatever, and then like, right, don't really do anything revolutionary in exchange. Convex is actually doing really revolutionary things. while kind of sticking to the rules that physics is leaving us with, um, for databases. So like, stuff like that's really cool. But yeah, like React, that came from— we were— we all went for like a dinner at one of my favorite restaurants the other night to kind of celebrate a bit, and we were all sat around a round table, and then, you know, there was an extra chair, so we moved the chair and we had to shuffle around the clock. And so I was like, ah, um, and yeah, I just— I remembered it. But yeah, there's like so many, uh, so many fun— There's still some that are going, you know, like that have like they were like the hot database of the day, um, but still like gained a user base and are ticking along really nicely.

Yeah, yeah. I just remember, I think me and Adam were talking about this where there was just a time where everything just felt a little bit more wild and things have settled down now and there's like tried and true ways that make sense for most people. But yeah, I do remember when like I was using like a different graph database every month and graph databases like blew up for a while. But yeah, it was good times and I'm glad I was able to kind of live through that.

Yeah. Now all that kind of frenetic wildness seems to just be marketing budgets, which is kind of not All right, now it's fun. You know, I tweeted this. I felt bad doing it, but I kind of— I've gained more conviction, which is there was this era of tech where basically if you were an engineer, the people— engineers were very hostile to anyone that like wasn't an engineer who couldn't code. And it was too extreme back then. I actually think we've gone too far the other way now.

Yeah, I saw that post. I agree.

Yeah, yeah. There's just people that are yapping 24/7 on Twitter that just don't know how computers work and have massive audiences, and it's just like You know what I mean?

Like, why?

I mean, I'm fat. Why would I— I wouldn't just sit there as someone who can't run going, here's running tips, here's how you think about running.

And then get offended when someone is like, do you actually run? You know?

Yeah, right, exactly. You can't say that. I mean, that's really, um, it's not really nice to say, you know, it's like, well, we're here to do software engineering, you know what I mean?

It's like, yeah.

If you applied this to other industries, it would be absolutely nuts, you know?

I mean, I do wonder though, like, I wonder because I'm like, I'm so in our industry, but I'm like, people should be the same everywhere. Like, are there— are these same dynamics, they exist in all industries and we just don't, don't see it? Like, are there medical people that are just like talking shit?

Yeah, like, I really hope not. No, I don't think they are. I really don't think there's like doctors that are just like there for the vibes, you know, and just like I hope so. I truly hope. I don't know. We'll see. But yeah, I mean, maybe that will change. I don't know. But it's definitely been a weird one to watch. The industry's in a very strange place right now.

Well, thanks for joining me today. This was a good conversation.

Thank you for having me.

Excited for you and your team and PlanScale Postgres. I always see pictures of your office. It looks really nice. I'll swing by at some point once I'm traveling.

Yeah, come through. Come and work from the office while you're here. People do that. It's really, it's nice to have people stop in. It's nice to have a physical location that people come to. We had this like really fun party the other night that we would just invite people to come and migrate. And just like being here at midnight, watching people just migrating their databases to PlanetScale while we were like hanging out, drinking, listening to music was just really fun. Like really enjoyable. SF's good for that, you know?

Yeah, that's the stuff I miss. I miss, I wish that I, that's what I'm jealous of. Like it's fun to have those things just happening randomly. On a weekly basis.

Yes. And you can actually, you know, like pretty much every day of the week in this town, there's 3 events you could go to. And there's interesting people everywhere. Then there's the other balance of there's like certain people or certain companies that go to every single one and you just think you can't be doing any work. You know what I mean? Like, I try and limit myself to going and doing something like that maybe once a week or whatever. But there's people It's just like, how have you— why does your Partyful say 600 events attended? It's just like—

Yeah, I feel like this is the case everywhere where there's like an industry and like some large percentage of it is just people acting out like they're doing stuff, and like a small percentage of it is actually doing stuff. And SF is like, no, like doesn't escape that either. And maybe it's even more concentrated in some ways.

Oh, for sure, for sure. And there's an inverse correlation with like People's pop— you meet people that are popular on Twitter and then you like talk to them about tech and it's just like, wow, you don't actually know how computers work. That's really interesting. Yeah, it makes sense. Why would one lead to another? You know, I don't know.

Yeah. All right, well, I'll let you go. Thanks for joining again.

Thank you for having me, as always.