Algora | Algora TV

Glauber Costa, founder of Turso - Open source interviews Ep.34

Algora

Watching now

00:00 Hey everyone, welcome to the Open Source Founder Podcast.

00:03 Joining me today is Glober, founder of Turso.

00:06 Thank you so much for joining us, Glober.

00:08 I'm super excited to hear about your background

00:10 and all the journey that led to the creation of Turso today.

00:14 Oh man, it's my pleasure. Super good to be here.

00:18 Thank you.

00:19 So, what was your life before this project looking like

00:24 and then what was the impetus that led to the creation?

00:28 My life, yeah.

00:29 Should I start from the beginning or do you want to know my life immediately?

00:33 Whatever you would like to say.

00:35 Both are interesting talk points, I guess.

00:38 But I wasn't really in the beginning of my career.

00:43 I wasn't really that interested in startups.

00:47 It's just not something.

00:48 I think I had some interest in like companies and et cetera.

00:53 But by and large, and keep in mind that this was before the crisis.

00:58 I'm old.

00:59 I mean, I shaved so you can't see, you know, the gray hair.

01:02 But like this was before the year 2000.

01:04 So like I was in was before the year 2000.

01:07 So I was in school around the year 2000.

01:14 So before the 2001 crash, which we now see a lot of the same pains and a lot of that.

01:16 But I was in school.

01:19 And in fact, I was not even going to school for computer science.

01:23 I was doing something else.

01:24 I eventually later gave up, changed, and graduated in computer science.

01:29 But I just took a liking to open source.

01:32 I mean, I, you know, just initially, I mean, I was a young guy.

01:36 Initially was mostly just out of like an ideological position, which, you know, just is a lot softer these days.

01:45 I mean, you know,

01:46 just that itself,

01:47 I think it's a podcast

01:48 that we can talk exclusively about that.

01:51 But also on the technical level,

01:54 I took a liking,

01:55 I took a liking to the Linux kernel.

01:58 So I still remember, for example,

02:00 one of the things that I,

02:01 when I was in this mental process of like,

02:03 do I switch courses or, you know, or not?

02:06 Like I was looking at the computer science program and they had a couple of courses on computer architecture, like assembly programming.

02:16 And I decided, hey, like, just let me see.

02:18 And I'll take a course, no credits.

02:21 I just sit there and see.

02:23 And then like I was using linux because of the

02:26 whole thing about like open source i you know none of us wanted to use windows at the time

02:30 um and and man it was just fascinating that this super low level stuff that most people do not know

02:38 um was just there in the kernel in front of me for me to check. So I started taking liking and I think that was around

02:48 2003. I started putting more and more time into, now let me just go look in the kernel and reading

02:57 books and et cetera. And I think around 2004, I sent my first open source contribution to the

03:02 Linux kernel. A little bit later, I was hired by Red source contribution to the Linux kernel.

03:07 Later, a little bit later, I was hired by Red Hat.

03:09 So that's how my career actually started.

03:14 And again, this is already super cool because I was not done with school.

03:17 I mean, I was still two years away, especially because of my switch.

03:19 Got a little bit late with that.

03:29 But I was working for Red Hat, you know, just to and trying my best to finish my studies, which I did.

03:31 So, I mean, the beginning was this for me.

03:32 I mean, just super.

03:44 My goal was like, how do I become as best as I can as a programmer, which I believe I mostly failed, just in some sense.

03:47 I mean, but like, how do I get you like the upper echelons of this?

03:50 And I just did this for 10 years.

03:55 I mean, it wasn't, it wasn't like I wasn't, I didn't really have, and come to think about it, like it's a terrible way to, if you want to go do startups, it's a terrible way to

04:00 start because like lots of the technical work in startups are not at that level

04:06 i mean they're like i think it would have been a lot uh you know if startups were my goal i would

04:11 have been a lot better served by like go at the time like the the big thing that everybody was

04:15 talking about was the rise of mobile right so just to go look into that i never wrote a mobile

04:19 application in my life i have no idea uh how to even write like a competent backend application, let alone front

04:26 end. And again, if I wanted to go in that direction, like building a company, maybe those

04:30 were things that were, you know, I would be better off. But that's not what I like to do.

04:35 My passion was always like, hey, like a super low level and let me see what this thing is doing

04:41 under the hood. And at Red Hat, Red Hat bought a company.

04:47 And that company was the company that was producing the KVM hypervisor,

04:52 which at the time, just to contextualize some of our listeners,

04:57 the big thing at the edge of the industry was the rise of cloud computing,

05:01 which everybody takes for granted today.

05:03 But at the time, it was something new.

05:05 And the technology trend powering this was the rise of virtualization.

05:10 So we had VMware as a company starting to do this.

05:13 But then Zen popped up as an open source company.

05:17 And again, KVM shortly after.

05:20 Red Hat acquired the company.

05:21 So I came to be very good friends with the people that were doing that company.

05:26 A couple of folks out of Israel.

05:28 Again, they worked with us inside Red Hat after the acquisition for a couple of years.

05:34 And then in 2013, I joined their startup as employee number three.

05:38 So that was my first experience as like a startup person.

05:42 I stayed with them for almost eight years.

05:45 So my career doesn't have a lot of people today,

05:47 one year here, two years there,

05:49 like interviewing all the time,

05:50 which again has some benefits,

05:52 I think, especially financially.

05:54 I didn't get a lot of that upside.

05:57 But in terms of like deep focus

05:59 and like really learning something truly well,

06:01 I mean, I think what I did works to some degree,

06:04 which is like,

06:08 you really need to commit to something and do that something for a long time.

06:14 And that was not initially a database company, but it pivoted into a database company.

06:20 So I already saw the pains of like, oh, what do you do when your product is not working?

06:25 So we pivoted into a database company called later renamed Scylla.

06:29 And again, that's where I stayed for almost eight years.

06:33 My co-founder was also in the Linux kernel with me.

06:36 He joined Scylla as employee number four right after me.

06:41 And he also stayed there for around eight years. And after that, I had a very short transition

06:45 in which that was the beginning of COVID and et cetera.

06:49 So I joined Datadog where I intended to stay for much longer,

06:54 but having money was so easy.

06:57 I did not, again, and it's funny because I did not know that.

07:00 For me, it was just like, that was my first foray.

07:03 So for me, it was just like the way things were.

07:05 The money was so easy that I had a couple of investors.

07:09 We had nothing.

07:10 We had essentially like it was my co-founder and I just wanted to start a company.

07:15 And we had investors already like ready to invest.

07:19 Again, we do have a reputation and all, but we didn't have a product or anything like that.

07:23 And despite, again, despite my intention being staying at Datadog for a lot longer, I just saw the opportunity.

07:29 The money was there.

07:30 And I said, hey, look, let's do it.

07:32 And here we are.

07:34 Thanks so much for sharing all this information about your background.

07:39 It's a unique background.

07:40 And you're right.

07:41 Nowadays, a lot of people jump between different companies.

07:45 Not just nowadays. I mean, back then too. I mean, the people who are chasing just the career part

07:49 of it and the money part of it, and not necessarily the subject, potentially think the best strategy

07:55 is to jump around. I'm not saying that my strategy was optimal. It definitely wasn't

07:59 from the financial point of view. And while you were there,

08:05 like these last,

08:07 these eight years,

08:08 were you considering

08:09 sort of like,

08:10 oh, in the future,

08:11 I would like to

08:12 strike my own path,

08:13 become an entrepreneur.

08:15 And then all the way

08:16 to actually, you know,

08:17 making that transition.

08:18 Was it a problem

08:19 that you faced on the job?

08:21 Was the excitement

08:22 with the technology aspect

08:23 that led you to

08:24 the creation of the mission? Oh yeah, I always wanted

08:27 and so even at Scylla, when Scylla started

08:32 again, I don't think it was, here's the thing, I'm the kind of person

08:35 that I want lots of things. Even

08:39 when I started in the Linux kernel, I was very interested at the Linux kernel, but

08:43 it wasn't the only thing I was interested at. I had like, maybe narrowing today, but especially as my younger

08:51 self, I was interested in many things at the same time. So there was always some level of opportunity

08:58 cost for me that I was, oh, what about this? Someday we'd like to do this. I love compilers,

09:03 by the way. So I wanted to work a little bit in compilers as well,

09:05 which I never did.

09:07 But I always had this sense of like,

09:08 oh, I could be doing something else.

09:10 And one of the things that I kept thinking,

09:12 hey, one day I would like to do was get my own company.

09:17 So at Seal, I tried my best to get as close as possible,

09:22 get their first customer.

09:23 I said, look, F this. I don't want to code anymore.

09:27 I mean, although I kept coding,

09:28 but I said, I don't want this to be my main function.

09:30 I want to go and try to help you win this customer.

09:34 And then the marketing folks would love me

09:37 because they wanted all this collateral

09:39 to talk about the product.

09:40 And I wanted, and I said, hey, awesome.

09:42 Let me write some content.

09:43 And like, and then the sales people love me because, uh, you know, they were having problems closing a deal.

09:48 And the buyer of databases are very technical. You know, that was true at Scylla as well.

09:53 Uh, it was very hard to get an engineer on the phone, but I was happy to do it. I was actually

09:57 annoyed when I wasn't called. Uh, and, and the CEO was like, dude, no, you cannot go in all of

10:03 those calls. Uh, you know, you have to protect your time.

10:06 Your time is very valuable to us.

10:08 I mean, those salespeople, they see you willing to help, they abuse you.

10:12 And I said, but for me, it was a learning experience.

10:14 I loved it, right?

10:14 I wanted to do that.

10:16 So we actually fought because I wanted to do more of that.

10:20 And, you know, just so I wanted that experience.

10:23 I wanted to see how it was going to be.

10:25 But my original plan was to stay like for four years a day to put my life in order because, dude, between us, I never made money in my life.

10:32 Just that's the thing.

10:33 And joining a public company at the time, you know, at Red Hat, I had no shares because when I joined, I was super young and et cetera.

10:43 Like I wasn't all I wanted. Like I was still hacking the Linux cetera. Like I wasn't, all I wanted,

10:46 like I was still hacking the Linux kernel.

10:47 That's all I wanted.

10:48 That's it.

10:50 And again, at Scylla, Scylla is still not,

10:52 didn't have an exit, you know,

10:54 and it's going to become even more challenging now

10:57 with the current environment.

10:59 So again, I was always making like a fair amount of money

11:03 for like as any technologist will do, but never like anything like life changing.

11:11 So a data dog for me was, you know, especially I have three kids now.

11:14 I had one at the time, but it was my time to like put the house in order.

11:20 But man, when you got the bug, you got the bug.

11:21 I saw the opportunity.

11:25 And then one year later, I was already out.

11:27 So I had one year in my life in which I made money.

11:29 That's it.

11:31 Well, thanks so much for being transparent about this.

11:33 And not many people say it outright.

11:36 Yeah, lots of people don't know.

11:37 They think, oh, you know, just go fight.

11:39 But the thing is like your cash compensation at Big Tech.

11:44 And look, startup is going to be a little bit lower.

11:47 So my take-home pay today as the CEO of the company

11:52 is lower than at Datadog, right?

11:56 But it's actually not that much lower.

11:59 It's actually lower than even Scylla,

12:02 but it's not that much lower.

12:04 So it's a little bit lower than what I was making at Scylla as an employee.

12:09 So if your listeners are thinking about starting a startup, that is the first thing you need to know.

12:15 Like your cash competition will decrease.

12:20 But that's not the worst part, because if you are working for a public company like Datadog was, and again, that is a part of the reason why I joined, you have stock at the company, which again, you also do at a startup.

12:31 Startup stock is very illiquid, which was why most founders don't talk about themselves as being poor.

12:39 I'm not poor.

12:40 I am illiquid.

12:42 Right.

12:42 Illiquid is just a person who is very rich on paper.

12:47 So startup stock is like this.

12:49 I mean, it's worth some dollar amount, but you can't touch it.

12:54 And the valuation is mostly horse crap because it could be worth 10 times actually more than that in the future.

13:00 It could be worth zero or maybe you'll never exit.

13:04 But if you go work for a company like Datadog,

13:07 you go check the ticker today, and I can tell you the price. So that's it. I mean, in every

13:11 quarter, you get that money, right? So it was very refreshing,

13:16 and also, I mean, it was not just the money, but I think the money

13:19 after 20 years was a... 20 years of getting better

13:24 as an engineer, but not making money,

13:27 my original plan was to,

13:29 hey, 40 years and I can't pay off my house.

13:31 I don't live in that expensive location anyway.

13:34 I put my life in order.

13:36 But the difference, as I was saying,

13:39 between Scylla and today is not that high.

13:42 It's like a delta.

13:43 But if you count the difference

13:44 between that year that I had

13:45 a data dog and now,

13:48 it's ridiculous.

13:50 Especially because stock

13:52 in 2020 had

13:53 only one direction it would go.

13:56 Right up.

13:57 So you sign your contract and then one

13:59 year later, you're already making a ridiculous

14:01 amount of money compared to you even thought you

14:03 were going to make when you signed the contract. The silver lining is that I don't think I will

14:09 be making that much money today because then data dog laws have its value. But look, man,

14:13 it is the life. In many ways, I think I'm better now because I had a life experience that I wanted

14:21 and I would not have had at the public companies. It's all about the experience versus...

14:27 You can't look at one factor, but it is a reality.

14:30 It is a reality that you're leaving a lot on the table if you come this route.

14:34 But I think this is helpful for people to hear, especially thinking about starting a

14:38 company.

14:39 And so I appreciate laying it out for people.

14:43 Now for the engineers, for the technology crowd,

14:45 transitioning the topic,

14:47 maybe you could tell us about TUSHO, what it is,

14:49 and then what 2023 looked like

14:51 in terms of the traction of the project, the development.

14:54 And then we'll get more into what's coming up next.

14:59 So you also asked me a question in the beginning,

15:01 like why did you start this company and et cetera.

15:04 And I talked about why I started a company.

15:08 But the thing is that the company was also not doing Turso in the beginning.

15:13 So we did not start Turso.

15:14 We started something else.

15:15 Becca and I started something else.

15:18 And it's funny because if you would ask them, if you have asked me when I joined Scylla, join a database company,

15:26 my answer would probably have been no.

15:28 Why would I do that? Sounds stupid.

15:31 But then after

15:32 eight years at a database company, I told

15:34 Pekka, hey Pekka, let's start a company.

15:36 What do you want to do? How about we do

15:38 anything but a database company?

15:40 Because again, why would we do,

15:42 why would we be so sadistic

15:44 and if you're masochistic,

15:46 if you're talking about yourself,

15:49 to start another database company?

15:50 So that's not what we did.

15:52 The thing is, man, just again, fate is fate, life is life.

15:56 And we released in October or November 2022.

16:03 We had our previous product.

16:06 And then we released, in the context of that product,

16:09 like we were using Cicolite, we were using Cicolite a lot.

16:11 And in the context of that product, we released a fork of Cicolite.

16:17 And this is on our website for you to read.

16:19 I mean, our reasoning, our motivation, and et cetera.

16:22 Again, happy to expand on that, but just to keep the flow of the story going,

16:26 we put this fork out.

16:29 In five days, we had 50% more GitHub stars on that fork than we had into our

16:36 product that we were building for a year.

16:39 But that's not all.

16:41 That fork had no code.

16:44 Originally, and we were lambasted on

16:47 hacker news because of that but it was a conscious decision like again we strategically decided to do

16:54 that i in hindsight i think it was the right decision obviously like uh could have been the

16:59 wrong decision as with any decision but we decided to announce the fork of SQLite, put the repository out there, but not write a single line of code.

17:09 All we did was we put a readme with our intentions,

17:12 like this is why we did it.

17:14 We put the manifesto out.

17:16 We put like a license because SQLite has no license.

17:18 It's something that lots of people do not know.

17:22 A code of conduct and et cetera.

17:23 And they said, hey, we want to build a community with this.

17:25 This is why we're doing this.

17:27 Come build with us.

17:28 And then again, the reaction for a lot of people in Hacker News was like,

17:33 hey, look, I mean, who are those guys?

17:35 Why are they doing this?

17:36 I mean, talk is cheap.

17:39 Where's the code?

17:40 I mean, you're just clowning around.

17:42 Where's the code?

17:42 Blah, blah, blah.

17:43 It was funny because somebody said, I trust the SQLite guy a lot more than a random stranger on the internet.

17:51 And then this dude that worked with me at Red Hat came to my defense and said, hey, look, Glover is not a rando.

17:58 I mean, look at the guy.

17:59 I mean, just that he's done this and this and this and that.

18:02 But from the point of view of the usual arbitrary hacker news person,

18:07 just a rando saying a bunch of stuff, but where's the code?

18:11 We did this very, it was very strategic that we wanted to do this this way.

18:17 Because?

18:18 What was the core idea behind this strategy?

18:20 Was it to see if people want to be part of this new journey?

18:25 No, because we understood that we...

18:28 So we had the discussion.

18:29 First of all, we knew that the reality,

18:33 I mean, in other situations,

18:34 the reality is different.

18:35 Like if we were just...

18:36 If we had this as a project,

18:38 as a conceptual, purely open-source project at home,

18:41 as a side gig,

18:43 it would have been a different story.

18:44 But we knew that for better or worse,

18:47 this was in the context of our business.

18:49 We did not have the necessarily the intention to pivot,

18:52 but this was a business decision,

18:54 like forking this or something that said we're going to do this because it

18:57 makes sense for us to do.

18:59 Obviously we knew that it could happen,

19:01 that they would catch fire as it did,

19:02 but we were not necessarily like expecting, like we knew there were scenarios in which this would could happen, that it would catch fire as it did, but we were not necessarily expecting.

19:07 We knew there were scenarios in which this would not happen.

19:11 And so one of the things we wanted to fix with SQLite

19:15 was this idea that open source but not open contribution.

19:18 There was a lot of chatter online about, hey, SQLite could be so much more

19:23 if only this, this, this, and that.

19:26 So we wanted to fix that social problem first and foremost. You know, fix it in the open source way,

19:32 which is like, you don't like it, fork it, which is what people say. And we don't have access to

19:37 this community. So we will do what was done before and like succeed or not, that's our attempt.

19:44 But our, so there were two considerations.

19:45 One of them is that we do not want to take two years

19:48 to put this out.

19:49 Because if we were to code all the things we wanted,

19:54 it would take a year, two years, right?

19:57 And then we would present to the world

19:58 this fork of SQLite that does all of the kind of stuff.

20:02 But one year later, two years later, and then what?

20:04 Then we run out of money.

20:05 The company doesn't exist anymore,

20:07 especially because we were already doing a product for a full year.

20:13 But the other reason is that we knew that, okay,

20:18 because another path is that you can say,

20:19 what is the most impactful thing you want to do?

20:21 And then you do that most impactful thing kind of willy-nilly,

20:24 not necessarily to the end,

20:25 but then you use that to show what you came for.

20:29 But our fear is that we would start then to be known

20:33 as the people that are doing SQLite plus that,

20:37 whatever that will be, right?

20:39 Oh, you know, we had this,

20:41 we run this internally in like those simulation scenarios with ourselves.

20:46 And you can easily see people chatting about us.

20:50 Hey, have you seen the guys who've done SQLite plus HTTP?

20:53 Or have you seen the guys who've done SQLite?

20:55 This technical thing that you do becomes the center.

21:00 So we decided not to write any code so that there wouldn't be a center for the conversation.

21:06 The conversation would be around the license and the community and those aspects, which kind of worked because a couple of weeks later, people were referring to us on Twitter as the fork of Sikolite.

21:18 And that's exactly what we wanted.

21:20 So it paid off.

21:21 But again, it's risky because we had to start producing colica very soon.

21:28 And we had this initial wave of

21:31 reaction of people just saying, hey, those guys are just a poor day.

21:35 But it's also not that bad because every time you have people

21:38 doubting you and then you can quickly prove that

21:43 hey, it is for real, it's great.

21:46 It's the best way to create believers.

21:49 We had that at Scylla as well.

21:50 Scylla had some very outrageous claims that, oh, we are 10 times faster than a Patrick

21:55 Cassandra, which was the time, the competition that we had.

21:58 And every time somebody called us on it and said, I don't believe you, and that's an

22:03 outrageous claim, that was great because they would run it and it will be true.

22:07 And then at that point, you become a true believer, right?

22:09 When you had originally the doubt.

22:12 So it's all good.

22:13 I mean, it worked well.

22:16 And then because we saw so much interest in such a short period of time,

22:22 we decided around end of November

22:25 to just kill the previous product.

22:27 End of November and early December 2022,

22:31 this is all fuzzy to me.

22:33 I was having my third kid at the time,

22:36 so just to spend some time in the hospital, et cetera.

22:38 So the timelines are a little bit fuzzy.

22:40 But around December 2022, we just decided,

22:43 hey, look, I'm out with this.

22:46 And we didn't even fully stop working on the previous product.

22:49 We kept a little bit because we had this idea that, okay,

22:54 just the fork of SQLite is not a business.

22:56 So do we have a business here?

22:58 So we worked for like another two months.

23:00 And then in February, we decided to launch the private beta of Thurso.

23:04 And then seeing the

23:05 reaction to that and say, okay, that's a business. Obviously not at that point, but this can become

23:11 a business. So now, kill the previous thing, do the new thing, and that's it. We had this transition

23:17 period. And the idea came during that period, sort of like November, December 22 to February 23,

23:23 or you were already kind of like thinking about

23:26 okay what else can we build but uh yeah no pretty pretty much yeah pretty much because uh again we

23:32 saw i think obviously again our previous product was also based on cyclite uh and so we had some

23:41 exposure to that and i i actually think that Turso, initially the way it was,

23:47 was very informed for better or worse by the things we were doing right before.

23:52 So we were trying to essentially was a global state management API for edge

23:59 compute.

24:00 And I think the reason, and we wanted to use SQLite for that because we knew that SQLite would, if done correctly, would replicate very well.

24:10 And what I mean by very well is that it's very cheap to run a set of like 10 replicas or anything like that.

24:17 And in the beginning, I think we focused a lot on Edge.

24:20 Today, we focus a lot less on Edge.

24:23 And happy to share why, because after a couple of months that we were like public beta and the product that was released and people were using, we started noticing the 70% of the people who were creating databases on Torso did not create a single replica.

24:43 who were creating databases on Tor, so did not create a single replica.

24:50 And there are two reasons for that. First of all, is that I particularly think that the word edge was a little bit butchered by Vercel, Cloudflare and others. The JavaScript community, and I was

24:56 not a part of that community originally, nor do I consider myself a part of this community today,

25:00 but it is a community that we serve a lot.

25:10 Lots of people in that community have a very twisted understanding of what edge means.

25:14 And then you hear things like, I don't like to use edge because it's not compatible with Node.js.

25:16 And then I say, look, edge is, for me, a geographical statement.

25:20 It has nothing to do with any runtime.

25:22 But again, because of specific technology and go-to-market decisions that those companies had,

25:28 I think mostly Vercel, in all fairness, people started associating

25:32 edge with certain restrictions that have nothing

25:36 to do with the replication thing. So people sometimes would come to Turso

25:39 looking for an edge database, but they would still not create replicas

25:44 because they didn't associate edge with replication.

25:46 They associated edge with HTTP

25:50 instead of TCP for database connections

25:53 and no need for connection pooling

25:55 and code starts and serverless

25:57 and things like that, which are accidentally edge

26:00 or not edge at all, if you ask me.

26:04 But also because we started noticing that a lot of people just wanted a way to run SQLite

26:12 in the cloud, right?

26:13 Just a lot of those other 70%, they were very interested in our product.

26:18 And we know that because then we started asking them, okay, so you did not create any replica.

26:21 What are you doing here?

26:23 What brings you here?

26:25 You invest this in the best way possible.

26:26 What brings you here?

26:28 And then you would hear things that have nothing to do with,

26:32 like, oh, because I want to run this in six regions

26:34 or ten regions or anything like that, right?

26:36 So over time, again, we expanded the scope of what we do.

26:41 But I think the reason we started focusing so much on edge

26:44 is because that was already what we were. But I think the reason we started focusing so much on edge is because that

26:45 was already what we were trying to do. Right.

26:47 So it kind of informs you again,

26:50 we do not start with a clean slate and maybe we should have,

26:54 we should have spent more time just thinking,

26:56 but it's hard as humans should, you know, just,

26:59 it's very hard to do this. You're all, you always get,

27:02 you always get, I think a little bit informed by your context,

27:06 by your surroundings, and your short-term history.

27:10 So that's how we ended up.

27:12 Again, so today I think I don't even talk anymore about Edge database.

27:18 Durso does replication very well, and it's still one of the things that we do.

27:22 But the word Edge is just quite problematic, so just don't talk about it.

27:25 And the product, I mean, does a lot more

27:27 other things than just that today.

27:31 I understand.

27:32 No, I understand that these things have been

27:33 conflated. And so what is the best way

27:35 to actually describe to people today what

27:37 TourStory is and where it's going

27:39 now in 2024?

27:41 SQLite and production is the best way.

27:44 So when you go look in...

27:49 My co-founder actually likes SQLite for modern apps.

27:52 But it's one of those phrases that is used so much

27:56 that it kind of loses its uniqueness.

27:58 It's not wrong, but I just don't talk too much like that

28:02 because of this, because it's an overused trope.

28:06 But there's a lot of people understanding and realizing that the time for SQLite has come for a variety of reasons.

28:15 And I think the reason number one is that machines are so powerful today.

28:20 But the workloads didn't grow that much.

28:22 But the workloads didn't grow that much.

28:30 So like I remember I come from this background at Scylla where we're talking about petabytes of data and billions of queries a second and et cetera.

28:38 But when you go talk to most companies, like if they have a lot of data, they will say, can Thurston handle my data?

28:39 I have a lot of data.

28:40 How much data do you have?

28:41 Like I have like hundreds of gigabytes.

28:43 There's not a lot of data.

28:45 I mean, it was a lot of data 10 years ago. Now it's not. Objectively, it's not a lot of data. How much do you need to...

28:50 And then say, I've heard SQLite is not great for writes. We hear this all the time. I heard

28:53 SQLite is not great for writes. Can you handle my workload? My workload is write heavy. Oh,

28:58 awesome. Tell me more. Let's put this in numbers. I write to the database almost every second.

29:00 Tell me more.

29:02 Let's put this in numbers.

29:04 I write to the database almost every second.

29:07 Again, this is not write heavy.

29:11 The write heavy workloads, the real write heavy workloads,

29:14 if you make that statement as an industry-wide statement,

29:18 stuff like we did at Cilla or log ingesting or anything like that, you're writing to the database hundreds of millions per second.

29:23 So if you're writing to the database every second,

29:27 it's not heavy writes.

29:28 You're right, SQLite is not good for heavy writes,

29:30 but we're talking about the fringe of the industry heavy writes.

29:33 We're not talking about the center of the spectrum heavy writes.

29:37 So SQLite is good for a variety of use cases today.

29:42 And it's extremely easy.

29:44 It's probably the best developer experience you've ever had

29:46 because you can start with just a file.

29:49 And it's extremely fast.

29:51 And because of that,

29:53 we can offer you a service that is extremely cheap

29:58 because of those things put together.

30:00 So there are many advantages of using SQLite as your main database.

30:03 But again, using SQLite in production is super hard because it's just a file.

30:07 So what do you do if you want to have another API server calling to the same database?

30:12 If it's just a file, right?

30:13 What do you do?

30:14 So there are many situations.

30:16 What do you do if you want to do backups?

30:17 What do you do if you want to scale this out?

30:20 What about fault tolerance if your API server dies?

30:23 What if your application is serverless? what if you are using, what if your

30:25 application is serverless? Where do you put the damn file, right? So what we want to do is like

30:30 plug the gap of those things. And again, we're doing this through the fork that we have. What

30:34 are the things that you can do to now make sure that SQLite can be your main database, can be used

30:40 for production, for your serious applications, like for the stuff that really moves the needle.

30:46 We believe that the answer is yes.

30:48 It can be used for that.

30:49 And we are the company that want to make it happen, right?

30:52 So again, just add just one thing, one of the things that I think we can do very well,

30:56 but there are others.

30:58 And look, if you go to our pricing page, you would see that in our most basic plan,

31:05 which is today the only one we have, but we're going to have others in the future,

31:09 for $29 a month, you can create 10,000 databases.

31:13 There's no other provider that can offer anything like that.

31:17 And you can only do this because of SQLite,

31:19 because every database is just a file.

31:21 There is no gigabyte granularity of creating databases.

31:26 There's nothing of the sort.

31:27 So you go and just want to have a database of 10 kilobytes,

31:30 have a database of 10 kilobytes.

31:31 And then if you want to have thousands of them,

31:33 you have thousands of them.

31:34 That's all fine, right?

31:35 And again, if those databases are not being used,

31:39 they don't consume any resources

31:40 because you don't have like a database server running.

31:43 We have like the one server serving all of those databases.

31:46 So you can create an offering that is very compelling,

31:49 I think, for a lot of users because of that.

31:52 So this is what we're trying to do.

31:54 And in 2024, I think we're going to do more of that.

31:56 I mean, we haven't reached yet the point

32:00 where we're monetizing this very aggressively.

32:03 So, I mean, as a company, we do need to survive.

32:05 We do need to find that path to monetization. And 2024 is the year in which I expect this to

32:12 happen. I expect us to be able to handle more and more critical. The database market, the good thing

32:17 about the database market is that it's very hard, very competitive, very cutthroat, but very easy

32:23 to understand. If you power a mission-critical workload, you make money.

32:26 That's it. Boom.

32:27 It's simple, right?

32:30 So that's where we're heading, hopefully, this year.

32:33 I love it.

32:34 And your work personally, I remember from your previous startup experience,

32:38 you love being customer-facing and doing sales, writing content,

32:43 helping market the business and, you know, bring new paying

32:46 customers in. So how

32:48 has, and that's an edge,

32:50 that's a superpower for a founder,

32:52 for a CEO to have. What does that look like

32:54 today? Do you get as much time as you

32:56 would like to on this sort of like

32:58 activity business function?

32:59 Yeah.

33:00 So the only thing I don't do anymore compared

33:03 to my previous life is code, unfortunately.

33:08 And it's funny because lots of founders keep coding.

33:12 Again, I had still coded a lot.

33:16 I don't know, man.

33:17 I'm older today.

33:19 I've got my family, my kids.

33:20 I don't have that much time outside of the business hours to be doing this.

33:26 But if you go look into the repositories that we have that are open source,

33:32 you will see my name on it.

33:34 But it will also notice that over time it drops precipitously.

33:40 And it's funny because I had this thinking when I was conceptualizing one day I'm going to have my company.

33:47 Will I code or not?

33:49 And at Chisel Strike, our previous product, I still coded a fair amount.

33:56 And again, we were a small team and et cetera.

33:58 I kept thinking, look, why would I not do this?

34:02 Because I can code and the company is not that big.

34:07 So this idea that, look, this person is not going to code at all

34:10 if you're a 10-person team,

34:12 immediately what you're saying is that you're losing 10% of your coding power.

34:16 So why would you not do this, right?

34:18 But as Thrusu started gaining muscle,

34:23 first of all, the amount of work that I have to

34:26 do increases.

34:28 As just one example,

34:30 we are pursuing SOC 2 certification now.

34:32 So the engineers do

34:33 have to get involved to make sure that everything's compliant.

34:36 But a lot of the work, man, is just like

34:37 reading stuff, creating

34:39 agreements, signing papers.

34:42 There's a lot of work that

34:43 I don't want my engineers to do. So I'll do it. I'm happy to do it. There's a lot of work that I don't want my engineers to do.

34:45 So I'll do it.

34:46 I'm happy to do it.

34:48 There is a lot of this kind of work of like talking to people,

34:51 coming into a great podcast like yours to talk about the product,

34:55 which is a marketing activity for me at the end of the day.

34:58 There is activity like going on LinkedIn and doing something I always

35:02 hated doing, like, which is just cold calling people.

35:06 You try to at least do a warm call,

35:09 but just to talk to people,

35:11 you get people to POC the product.

35:13 So there's a lot of that.

35:15 But also I started noticing that at some point,

35:18 and it was very visible,

35:19 like it was almost like physically visible.

35:22 I was doing more harm than good every time I coded

35:25 because you don't have the context.

35:28 Like you become like essentially a, look, yes,

35:31 the other day I found something, an error message

35:34 with the right word.

35:35 I fixed it, right?

35:37 But if you do any serious code, like you need context.

35:41 You need, coding is like, you need to sit down

35:43 and you need to understand what you're doing

35:44 and you need to understand the product. doing. And then you understand the product.

35:45 And I don't have this kind of time anymore

35:47 because I have like 30 minutes here,

35:50 one hour there.

35:51 So you can do more damage with that

35:53 than you can help

35:55 because like in 30 minutes,

35:56 you're not going to even have like a,

35:58 and then I have no time for two weeks.

36:00 And by those two weeks,

36:01 a lot of things have changed.

36:02 And so I just kind of decided that

36:06 I was almost always better off at this point, asking people from my team and making them know

36:13 that I consider that a priority, right? Hey, and why I wanted to do this. This is the reason I

36:20 don't have the time to do it. Can you do it? Then coding this myself. So this is the only thing I think

36:26 I don't do anymore.

36:28 And again, I don't have

36:29 any free time because

36:31 of family, so I just don't code at

36:34 all, which is a shame.

36:36 But the other

36:37 part of the work, yeah, absolutely, we do a lot

36:39 of that.

36:41 That's great. And of course,

36:42 the team team internal and

36:45 broad has

36:46 expanded to

36:46 help with

36:47 the distribution

36:48 and sort of

36:49 like marketing

36:50 motion.

36:51 And for that,

36:51 I know that

36:52 you've had a

36:52 very successful

36:53 partnership so

36:54 far with the

36:55 PrimeGen

36:55 and you're

36:56 planning to

36:57 help him

36:59 fly to

36:59 Brazil.

37:00 We will,

37:01 yes.

37:02 So I wish

37:04 I had the

37:04 final dates, but I don't we're uh we're about

37:08 to close the venue so when we close the venue like we have the tentative date which is may the 11th

37:13 but again we need to also find a venue that's going to be available on may the 11th and and

37:18 we haven't closed it yet there there's some uh but but uh yeah man that was that was a little

37:24 bit like that's how you end up doing things in startups.

37:26 I mean, you don't plan that much.

37:28 I mean, you see an opportunity.

37:29 You say, hey, what if?

37:32 So I noticed I was actually born in Brazil, although, again, I'm not in Brazil for many years.

37:38 It's going to be more than 15 years, actually, probably this year.

37:44 And I don't have a lot of context still in the country.

37:49 Like, I don't go much.

37:51 So just, but, you know, just I still know it.

37:57 And I know a lot of things about the country.

38:00 So I think I understand the culture,

38:02 I understand the things that people value to some extent.

38:06 And I saw that, like, every time that PrimaGen would mention Brazil

38:08 in his stream, people would go crazy

38:09 say hey come to Brazil, come to Brazil, come to Brazil

38:12 we would love to have you here

38:13 and again I've developed

38:15 as you mentioned a very good relationship

38:18 with him, he talked about us

38:20 on stream many times

38:21 and then one day he put a tweet out

38:24 that just had the word brazil

38:26 yeah and this is like the the most that guy's a genius man like this is just uh it takes it to

38:33 the next level like this thing about engagement just put the word out there and then he had 160,000

38:39 views at that moment i said not like i i asked him look if if i if i find a way for you to go to

38:46 brazil okay you want to go do you want to come and initially he was a bit resistant but i think

38:52 with with time over time he saw like how cool this could be uh so so we decided as soon as he gave me

38:59 the okay we said yeah let's just do it so we don don't have, again, I don't have any presence in Brazil,

39:09 and I don't have any contacts, any good contacts that could make it happen.

39:13 So we partner with one of our customers that is from Brazil,

39:15 which is a company we like a lot called Azion.

39:18 They do amazing work.

39:22 They do a lot of like edge compute as well, the real stuff edge compute.

39:26 So they have a compute platform like Cloudflare,

39:29 workers, like the stuff that Akamai has.

39:34 And I think you as the listener should just go and try.

39:37 And again, they are in Brazil.

39:38 So I called their CEO and said, hey, look,

39:41 you want to help me with this?

39:43 They said, yeah, sure, let's do it.

39:44 And that's how this whole thing was born.

39:46 I love it.

39:47 And you're going yourself then, I reckon.

39:48 I am. And most remote companies, it's very good for the culture that you meet at least once a year.

39:56 And we usually do this in Europe because it's cheaper.

39:58 We have a bunch of people in Europe.

40:02 So it's just cheaper to fly people to europe uh in our case

40:06 but uh we usually do this around april or may this year we were considering maybe not doing uh

40:11 just out of like we want to be responsible because we know the economic situation is really bad

40:16 however it does make a difference i mean every time we met before we could see like that there's

40:21 something very big happen like big lots of things that as a remote company,

40:25 sometimes are not going well when you meet,

40:27 like you kind of solve many problems at once. You know,

40:31 remote has many advantages, but also has some disadvantages.

40:34 I personally think both models can work.

40:36 You just have to be very cognizant about what you're doing. And again,

40:39 if you, if you, and, and, and do your best to, to patch the disadvantages.

40:44 And one way of doing this is just this, you have to meet.

40:47 So I was already like, look, financially speaking, we shouldn't meet.

40:52 But on the other hand, it's not that this is going to be like the make or break it for the company.

40:56 It's not like, oh, I haven't met this year.

40:59 Because of that, I had three more years of runway.

41:01 It's not like that.

41:02 It's not even a month of runway.

41:03 So it doesn't make that much difference.

41:06 It makes a very big difference in the culture of the company.

41:11 And flying everybody to Brazil is more expensive

41:14 than it would be to fly everybody to Europe.

41:19 But if you already have to fly me to Brazil anyway,

41:23 and we wanted our VP of marketing to come and et cetera,

41:26 at that point, it's not that much more expensive

41:29 to just do the whole thing there.

41:30 So I'm probably going to do this as well.

41:33 That's neat.

41:34 That's super nice.

41:35 Well, safe travels to everyone when the time comes.

41:37 I hope you have a great time.

41:38 Thank you.

41:38 We've got a lot of media around it too for us to follow.

41:42 Looking forward to announce the final date.

41:44 And for the people in your audience that are from Brazil,

41:48 look, we would love to have you there just hanging out with us

41:52 and learning more about Thrusa, learning more about ASEAN,

41:54 our partner, and interacting with the man, the myth,

41:57 the legend, the primogen.

42:00 Everyone's excited.

42:01 Thank you so much for this initiative.

42:03 It's going to be super cool.

42:04 And we didn't talk all too much about the open source motion and what the community has to say. Everyone's excited. Thank you so much for this initiative. It's going to be super cool.

42:11 And we didn't talk all too much about the open source motion and what the community has like, the contributors. And actually, perhaps we could couple this with, if you'd like, sharing your screen, showing us the repository and then transition to a demo.

42:20 This is our main repository. And we have the CLI, which is also open source.

42:24 We have other things that are also open source. We have other things that are also open source.

42:26 We have the clients that are also open source.

42:28 But when I mentioned earlier, like, that we wanted to fork SQLite and get it going, this is what we did.

42:33 Like, there's a project called LeapSQL.

42:36 And I wish I had the foresight to just call it Thurso.

42:40 But, again, Thurso, as I explained, does not exist.

42:42 but again,

42:42 Thurso,

42:43 as I explained,

42:43 did not exist.

42:45 We kind of did it like,

42:48 more or less,

42:50 I'm not going to say an accident,

42:51 but almost an accident.

42:54 So we have Leap Sequel.

42:55 Leap Sequel,

42:57 one of our goals was to have,

42:58 as I said,

43:01 to be the open contribution fork of SQLite,

43:03 be this thing that people can come in and build together with us.

43:04 So the thing that I love the most,

43:06 like the stars are great.

43:08 So we have 6,000 GitHub stars,

43:10 which is more than SQLite.

43:11 Although I am very cognizant

43:13 of the fact that SQLite

43:14 is not developed in GitHub.

43:16 It has a mirror in GitHub.

43:17 I think if it was developed in GitHub,

43:18 it would have more,

43:20 but still a great milestone.

43:22 But the thing I am proud the most

43:23 is just the number of contributors, right?

43:25 So if you look in here, we have 54 contributors to Leap Sequel.

43:30 And we don't have 54 people in this company, I can guarantee you.

43:34 There are, as usual with the open source, I mean, you have like your main people doing a lot of work.

43:41 And here's me.

43:42 But then you start having like people who have never been with this company.

43:46 And all of those, Doug was our previous dev rep,

43:50 but lots of those people are just people

43:52 who are super excited to contribute to the project.

43:56 And these contributions, they come in many ways.

44:00 Some of them are for the build system.

44:01 Some of them are for the core database.

44:03 Some of them are for the server.

44:05 So we have contributions in many places.

44:08 Again, here you're going to see all the description

44:12 of what is different between LibSQL and SQLite.

44:18 And LibSQL can also be run as SQLite as a pure file,

44:22 or it has a server that it can connect you and then send your SQLite

44:26 statements to us. All of that is in the same repository. But again, even if you run as just

44:31 SQLite, there are a couple of differences between LibSQL and SQLite, and they're listed here in

44:39 terms of feature. Durso is a service that we manage based on that.

44:46 So let me just show, you know what, let me just show our website.

44:52 Sounds good.

44:53 Going to come here.

44:56 Durso.tech.

44:58 We are, again,

44:59 we are reviewing a lot of that messaging to focus a lot more on this idea of

45:04 like a just SQLite. One of the things that

45:07 we're putting front and center now is just the insane amount of databases that we can create.

45:11 If you come here to our pricing page, you see we allow people to create up to 10,000 databases

45:16 already on $29. We're going to have a new plan coming up, hopefully at the end of the quarter,

45:23 with more stuff. And you can see you can do all of the end of the quarter with more stuff.

45:25 And you can see you can do all of the kind of things that are common in databases today.

45:30 You can do database branching.

45:32 You can access this from serverless.

45:34 You can do point-in-time restore, automatic backup.

45:37 So all of this kind of stuff that you will need to run a production database is done by us.

45:43 Now, if I share my terminal now, which is here.

45:51 Right, so first of all forgive me, I'm not a JavaScript person

45:55 at all, but I know this is the language that the most people are into. So I

46:07 So I wrote by copying an internal example here,

46:10 just an extremely simple example.

46:15 So forgive me for any non-standard JavaScript stuff here.

46:19 So you can come here and it just creates a client.

46:23 And this is a super stupid example that just sends a query.

46:29 Select one. The cool thing about it is that you see that there is a DB URI environment variable.

46:32 I'm going to do something like this,

46:34 DB URI equals file SQLite.db.

46:40 Let me just show you first that this file does not exist here.

46:43 All I have here is the index.

46:46 So if I come here and I do this, and then I do bun,

46:53 because I want to, despite my age, I want to sound a little cool,

46:57 so I'm going to use bun.

46:59 So you come here and you do index.

47:02 You can see that it created the file.

47:04 You're going to see here that it created the file. You're gonna see here that it created a file

47:05 and it did select one.

47:09 I can come here now and do SQLite 3, SQLite.tb,

47:14 create table podcasts,

47:18 name text, insert initial podcasts,

47:24 and I could come and replace this Text, insert, podcasts, food.

47:25 And I could come and replace this and then do the same thing, right?

47:30 Just if I do this now, you would see, like, the column name and the text

47:36 and then the values, et cetera.

47:37 So we have, like, this is all, like, again, this is using a file.

47:41 I'm not talking to Durso.

47:42 You don't have to be signed up to Durso. You can

47:45 just do it. And this is one of the beautiful things about SQLite. The driver itself is the

47:51 database. There is nothing. I haven't set up a database. I haven't put up a server. I have done

47:56 absolutely nothing. The database is now part of my code. It's a library. And I just called it,

48:02 and it's there. The database itself is a file.

48:05 Now, what I can do as well is I can do torso-db create.

48:10 I just want to move this out of my way here.

48:15 Torso-db create test.

48:19 So you're going to see that in actually fairly slowly,

48:24 there's something most database creation should not take that long.

48:29 It usually takes less than a second.

48:32 I just created a new database in Thruso, right?

48:35 So I can do the same thing.

48:37 I can come here, and then I can do Thruso DB shell test.

48:41 I'm going to create the same table, create table,

48:50 db shell test. I'm going to create the same table, create table podcasts, name text. And I'm going to do insert into podcasts, name values bar. Right. And then I can do a torso db show test and you see that there's a url here right so i i can use this url just to

49:10 keep things simple i'm not going to put in the environment variable i'm just going to copy and

49:15 then again because this is over http i also i'm going to do token so i can come here and then

49:20 again torso db tokens create test so i I'm gonna get this whole thing here.

49:27 And I had troubles in the past with the token

49:29 just not being copied.

49:32 I guess I could put on an environment variable,

49:34 but let's alt-token.

49:36 Hopefully this is correct.

49:38 If you don't have the, let's come with this.

49:41 And if I just run this again, bun-index.ts,

49:44 you will see that you now contacted this remote database.

49:48 And another thing that is super cool that Thrusa can do is they can

49:52 also replicate the database

49:55 from our server to your client.

50:00 So I can essentially come here and let's see if

50:03 we don't have any demo effect here, but I can essentially come here, and let's see if we don't have any demo effect here,

50:06 but I can just remove that file, rmsqlite.tb.

50:12 And then I can turn this URL into a sync URL.

50:20 And just in the URL, I'm going to say file on cast.tb.

50:30 And then the first thing I want to do here is maybe sync.

50:33 So it's super simple.

50:35 And hopefully we're actually even going to do away with the need of calling sync explicitly.

50:39 We want to transform this into a background operation.

50:43 I think we need an extra comma in the URL line, by the way.

50:47 What's up?

50:48 I think we need an extra comma in the URL.

50:52 Yeah.

50:53 I kind of hate this, by the way.

50:55 I never understood why is it that in JSON you don't have a comma in the last one.

51:00 And it's not that you can't have a comma.

51:02 You must not have a comma.

51:05 Right. That's a cool thing can't have a comma. You must not have a comma. Right.

51:05 That's a cool thing.

51:06 So if you see here, right, it runs a bunch of stuff.

51:13 I mean, just when you go up bar, and you will notice that this file is now here.

51:20 So this file was created.

51:21 I can come here, like SQLite 3 3 podcast.tv, select star from podcasts.

51:28 This is a local file, right?

51:30 So you can now have this will periodically, not the way I wrote,

51:36 because you have to, the right way is to put the sync call on a background thread.

51:40 And again, we want to automate this because it's one of the common complaints

51:43 that our community has.

51:45 But you can code it in a way

51:46 that you're reading from your local file every time.

51:49 So you're now reading in microseconds,

51:51 but every change that happens in the server,

51:55 like if you have multiple writers and et cetera,

51:57 like if you have multiple API servers,

51:59 everything that happens on the server,

52:01 you replicate your local file,

52:03 which is again,

52:04 something that

52:05 lots of people find mind-blowing because you don't have that capability in standard SQLite.

52:10 Not to mention, as I said, you can create 500 databases for free. So you can come here,

52:16 say create DB2, and there you go, another database, and DB3, DB4, and keep creating

52:23 databases as many as you want.

52:26 And all of that is something that you can create with an API.

52:29 So you can actually just call an API as well to create the database for you.

52:33 So you can have a system where a user logs in.

52:37 It's a new user.

52:38 You create a database just for that user,

52:39 and then the user can do whatever they want with this database.

52:44 Maybe you manage the schema, and then one of the things we're working on,

52:48 because we hear a lot from users, okay,

52:50 but what if I want to evolve the schema of all of those databases at the same time?

52:54 But there are many scenarios in which the user now, again,

52:59 you were controlling the schema only because you didn't want users

53:03 messing up with each other's database.

53:06 But there are many good scenarios in which,

53:08 since I have a database per user anyway, just let the user do SQL,

53:12 let the user control the schema, let the user do whatever they want.

53:15 So you can have a situation like that where you have thousands of –

53:18 we support millions of databases in the system itself.

53:22 So you can create like a database per tenant,

53:25 a database per session.

53:26 You can create a database in less than a second.

53:29 You can delete a database in less than a second.

53:31 So those things can be very ephemeral.

53:32 I can come and show you a little bit more of the CLI.

53:36 So DB list is going to show me a list of all my databases.

53:42 And then as I said, like just you can have as many of them as you want.

53:46 And we have a,

53:48 we have a web interface as well.

53:49 If you go to torso.tech slash app,

53:52 you have your visual management of the databases over there.

53:56 So this is torso.

53:59 Awesome.

54:00 Thank you so much for walking us through torso.

54:02 And like all of this was super amazing,

54:04 super cool.

54:06 I especially like how flexible

54:07 the database is since it's just

54:10 a file. You can have it on the

54:12 local device, on the remote, on

54:14 the edge, and you can replicate

54:16 stuff very easily because of

54:18 these features.

54:20 That was the plan.

54:23 Awesome.

54:25 Folks, it was a pleasure being with you. Loads. That was the plan. Awesome. Awesome.

54:26 Folks, it was a pleasure being with you.

54:29 Thank you so much for making time.

54:31 This was awesome.

54:31 Really appreciate it.

54:32 Thanks, Omer.

54:33 Any parting words?

54:34 Anything you want to say?

54:36 Anything I should say?

54:38 Yeah.

54:38 Any advice for other founders or open source maintainers?

54:43 Last words from you.

54:46 Gosh, advice.

54:49 Advice for other founders.

54:50 I hate advice, man.

54:52 I hate advice because you ask 10 people

54:56 and you hear 12 opinions.

54:58 Like the most challenging thing with advice

55:01 is picking which of them you're going to listen to.

55:08 But, and again, if you know, if you're skilled enough to figure out

55:13 which one you're going to listen to,

55:15 maybe you don't need the advice in the first place.

55:17 But there are like, I don't know, from a couple of things,

55:22 I think I wish I had done differently.

55:23 I don't know, from a couple of things I think I wish I had done differently.

55:34 I wish I, you know, as I said, although I don't see how we would have done it in the situation that we had with the constraints that we had and all that.

55:46 I think I wish I spent more time in the beginning understanding, like, what is the real power of a SQLite-based system that users want versus just going straight into the edge thing.

55:47 But then again, when we launched our private beta, we put a form out and asked people,

55:54 like, what are you interested in?

55:55 A lot of those people said edge.

55:58 But for us, it was like, yeah, we have validation.

56:07 like, yeah, we have validation. But I think today after the experience that I had with this startup and founding my first startup, I don't understand the word validation the same way I used to.

56:16 Validation, the way I understood was just something like the, hey, this person said this is cool.

56:22 like the, hey, this person said this is cool.

56:27 Like I had a bunch of users saying this and that.

56:30 I don't consider this validation today.

56:34 Like validation is something a lot stronger than that, right?

56:37 The real validation doesn't come until someone signs the check.

56:39 So I wish I had spent more time.

56:43 And first of all, I wish I knew this in a more internalized way,

56:48 as I know today.

56:50 But in hindsight, I wish I had spent more time

56:53 in this process of real validation.

56:56 So I think one advice that I have for founders