Supabase: Cash Does Not Equal Success

1. 0:00 – 1:56

   Paul's Origin Story

   1. SPSpeaker0:00

      [upbeat music] Today I'm joined by Paul Copplestone, the co-founder and CEO of Supabase. Supabase is a Summer 20 company, uh, who started as a open source Postgres as a service, and has now become one of the fastest growing dev tool in this AI era, to the point that their last round, uh, put them as a decacorn valuation. Thanks for joining us, Paul.

   2. PCPaul Copplestone0:26

      Happy to be here.

   3. SPSpeaker0:28

      Um, so maybe before diving into this latest news, uh, you started the company in 2020, but it's not your first company, right? I mean, you, like, created two others before?

   4. PCPaul Copplestone0:39

      Yeah. This is my third venture backed startup, so I've done a few other-

   5. SPSpeaker0:42

      So venture backed, so you've done even others.

   6. PCPaul Copplestone0:44

      Yeah, yeah. [laughs]

   7. SPSpeaker0:44

      How many total?

   8. PCPaul Copplestone0:45

      Uh, I don't know. Um-

   9. SPSpeaker0:47

      You don't count anymore

   10. PCPaul Copplestone0:47

       ... I also helped a lot of people, um, launch their startups because I just really enjoyed that space, and in some ways that kind of is the genesis of Supabase. I mean-

   11. SPSpeaker0:58

       Yeah

   12. PCPaul Copplestone0:58

       ... Supabase people build startups on top of it, so it's kind of the thing that I-

   13. SPSpeaker1:02

       Full cycle.

   14. PCPaul Copplestone1:02

       Yeah, it's full cycle. Um, but in terms of my own startups, yeah, we- I did two venture backed ones before this-

   15. SPSpeaker1:10

       Mm-hmm

   16. PCPaul Copplestone1:11

       ... um, uh, both in Southeast Asia.

   17. SPSpeaker1:13

       And both got acquired for a lot of money?

   18. PCPaul Copplestone1:16

       No, they weren't as successful as Supabase.

   19. SPSpeaker1:19

       Okay. [laughs]

   20. PCPaul Copplestone1:19

       Um, you know, I still learned a lot and got a lot out of them. Um, the first one-

   21. SPSpeaker1:25

       But, um, how much of these first experiences influenced, uh, what you are doing now?

   22. PCPaul Copplestone1:30

       Quite a lot. Quite a lot. I learned a lot.

   23. SPSpeaker1:32

       Mm-hmm.

   24. PCPaul Copplestone1:32

       I think that's the thing. Um, whether a startup succeeds or not-

   25. SPSpeaker1:36

       Mm-hmm

   26. PCPaul Copplestone1:36

       ... you... A- and in fact, in many ways, if they fail, you learn a lot from the failures themselves. So, um, I think I would not have been able to build Supabase like we have today if it wasn't for doing those two startups.

   27. SPSpeaker1:51

       Awesome. And I think that's h- also how you got the idea. Is that fair?

2. 1:56 – 3:11

   Building Software He Needed 15 Years Ago

   1. PCPaul Copplestone1:56

      Actually, the idea, the idea started a long time ago. When I first started building, um, a- and my career-

   2. SPSpeaker2:04

      Yeah

   3. PCPaul Copplestone2:04

      ... I was doing contracting and really focused on the database side of things-

   4. SPSpeaker2:09

      Mm

   5. PCPaul Copplestone2:09

      ... so I've been building with databases and building databases-

   6. SPSpeaker2:12

      Okay

   7. PCPaul Copplestone2:12

      ... since the, uh, since I was at university, actually. I started contracting and working immediately, and funnily enough, one of the first jobs that I ever had was to build a real time system on top of a, a database for farmers, 'cause I'm from New Zealand.

   8. SPSpeaker2:28

      Okay. [laughs]

   9. PCPaul Copplestone2:29

      But, um-

   10. SPSpeaker2:30

       So that makes-

   11. PCPaul Copplestone2:30

       But yeah, so again, it's kind of full circle now after it, it took what, 10, 15 years, and now I'm building a startup that kind of has the primitives that, that I needed at the time.

   12. SPSpeaker2:41

       So you're building software you wanted to exist back, like, whatever, 10 years, 15 years ago?

   13. PCPaul Copplestone2:47

       Yeah, yeah. Well, funnily enough, the, now that I think back to it, the technology existed, but not in a form that you'd use today. I was using something called ColdFusion, which is a-

   14. SPSpeaker2:56

       Oh, yeah

   15. PCPaul Copplestone2:56

       ... technology by Adobe. [laughs]

   16. SPSpeaker2:58

       I know that one. [laughs]

   17. PCPaul Copplestone2:58

       You know the one?

   18. SPSpeaker2:59

       Of course.

   19. PCPaul Copplestone2:59

       Yeah, with the Flash charts and things. So that was actually very, very impressive. Um, so they were leagues ahead of their time, but, um, unfortunately they didn't build on Postgres.

3. 3:11 – 4:17

   Why Bet on 30-Year-Old Postgres?

   1. SPSpeaker3:11

      Okay. Okay, so Postgres, like, actually when you started, uh, uh, Supabase, uh, Postgres was already, like, 30 years old or something.

   2. PCPaul Copplestone3:18

      Yeah.

   3. SPSpeaker3:19

      Like, why bet on an old technology?

   4. PCPaul Copplestone3:22

      So we knew we wanted to build a database company.

   5. SPSpeaker3:24

      Okay.

   6. PCPaul Copplestone3:25

      Um, we liked the problems that databases-

   7. SPSpeaker3:28

      Mm-hmm

   8. PCPaul Copplestone3:28

      ... presented, and big problems, big, hairy problems present good opportunities for a startup founder.

   9. SPSpeaker3:35

      Okay.

   10. PCPaul Copplestone3:36

       Um, and then why did we choose Postgres? Mainly because I monitored Hacker News a lot, and I could see just this, um, very nascent but growing, uh, groundswell of developers choosing Postgres.

   11. SPSpeaker3:49

       Mm.

   12. PCPaul Copplestone3:49

       And I see Hacker News really as a sort of, uh, early adopter mentality. And then as we've gone on throughout the years, Postgres has now become the most popular database. So it was a good thing.

   13. SPSpeaker4:01

       And was it because of you?

   14. PCPaul Copplestone4:02

       Well, well, we'd like to think that we contributed a little bit-

   15. SPSpeaker4:05

       [laughs]

   16. PCPaul Copplestone4:05

       ... but I think the 30 years of development from the community and all the systems around Postgres-

   17. SPSpeaker4:10

       Yeah

   18. PCPaul Copplestone4:10

       ... are, are really the main thing, and hopefully we just, uh, added, uh, a little bit of fuel to an already raging fire.

4. 4:17 – 7:17

   The Open Source Decision

   1. SPSpeaker4:17

      So, so you placed your bet on Postgres. You saw it like, uh, as a old but upcoming in some way.

   2. PCPaul Copplestone4:22

      Yeah.

   3. SPSpeaker4:22

      The right technology for the future.

   4. PCPaul Copplestone4:24

      Yeah.

   5. SPSpeaker4:24

      Um, and then you decided to be open source.

   6. PCPaul Copplestone4:27

      Yep.

   7. SPSpeaker4:27

      Um, what was the intention behind that?

   8. PCPaul Copplestone4:31

      There wasn't too much thought. I think, like-

   9. SPSpeaker4:33

      [laughs]

   10. PCPaul Copplestone4:35

       ... um, Ant and myself are very philosophically aligned to open source.

   11. SPSpeaker4:39

       Mm-hmm.

   12. PCPaul Copplestone4:39

       I, I really like, um, the community around open source, especially, you know, like Postgres. It's-

   13. SPSpeaker4:44

       Okay

   14. PCPaul Copplestone4:45

       ... it's a community that is very distributed. It... Postgres belongs to no one. And so in this way, you, it actually has some very interesting dynamics. Like, oh, not only is it 30 years old, but, um, I expect it to continue being popular for the next 30 years purely because no one company owns it.

   15. SPSpeaker5:08

       Mm-hmm.

   16. PCPaul Copplestone5:08

       And because of this, then it has some very interesting dynamics. Like, oh, because no one owns it, everyone can contribute to it. Because everyone can contribute to it, it makes it better. So you end up with this very interesting flywheel where all the hyperscalers have to offer Postgres, which makes it better, and then, you know, we as well will contribute and, and hopefully-

   17. SPSpeaker5:27

       So everyone is contributing to the success of Postgres.

   18. PCPaul Copplestone5:29

       Yeah. So, so then for us, like I really like that model where everyone can contribute and, and feel like it's, um, not just, um, kind of developed by a single company.

   19. SPSpeaker5:41

       Mm-hmm.

   20. PCPaul Copplestone5:41

       And this actually points towards one of the differences that we have as an open source company. We try very much to contribute and give back to existing open source tools rather than, say, build our own. Um- You know, a lot of database companies will come in and say, "Oh, I'll build a bigger, better database, and I'll make it look like Postgres." Our way of doing things is to build a ecosystem around-

   21. SPSpeaker6:06

       Mm-hmm

   22. PCPaul Copplestone6:07

       ... uh, Postgres, but find the tools that already exist if they, if possible, and contribute to them. If they don't exist, then we'll, we'll definitely build them. But when we can-

   23. SPSpeaker6:16

       And when you build them, you always build them open source to this day.

   24. PCPaul Copplestone6:19

       Completely open source.

   25. SPSpeaker6:20

       Right.

   26. PCPaul Copplestone6:21

       So all of our licenses, uh, are either MIT, Apache 2, or, uh, Postgres license.

   27. SPSpeaker6:26

       So it's real open source. It's not, uh-

   28. PCPaul Copplestone6:28

       Completely free, yeah

   29. SPSpeaker6:29

       ... kind of like a, yeah, that's open source

   30. PCPaul Copplestone6:30

       Can be used without restriction. Yeah.
5. 7:17 – 8:33

   Do Cloud Giants Threaten You?

   1. PCPaul Copplestone7:17

      of the tools.

   2. SPSpeaker7:17

      And you never worried about, uh, the big cloud providers to take o- like, pretty much make more money than you from the project by reselling Supabase, given it... because, like, the license lets them do that.

   3. PCPaul Copplestone7:29

      Yeah. I mean, that's always a topic of discussion, as you would know-

   4. SPSpeaker7:33

      Yeah

   5. PCPaul Copplestone7:33

      ... in, in boardrooms. Um, Supabase is a little bit different in that, first of all, we're going all in on Postgres.

   6. SPSpeaker7:40

      Mm-hmm.

   7. PCPaul Copplestone7:40

      And they already have their Postgres offerings. So in many ways, we're kind of competing with them rather than taking what are they going to-

   8. SPSpeaker7:47

      Because they have, like, their own version, own flavor.

   9. PCPaul Copplestone7:49

      Their own flavor of it.

   10. SPSpeaker7:51

       That would be-

   11. PCPaul Copplestone7:52

       Yeah

   12. SPSpeaker7:52

       ... that would be very weird for them to actually offer Supabase.

   13. PCPaul Copplestone7:54

       Take a, another one and offer it on top.

   14. SPSpeaker7:55

       Because then they would compete with themselves.

   15. PCPaul Copplestone7:56

       Yeah, exactly. And also, that points to another thing. Supabase is an amalgamation of a few tools that, um-

   16. SPSpeaker8:03

       Mm-hmm

   17. PCPaul Copplestone8:03

       ... centers around Postgres. So for them to spin this up, it wouldn't be like standing up a Redis instance or Elasticsearch or something like that. It would be more building an entire ecosystem around, um, and many-

   18. SPSpeaker8:16

       Mm-hmm

   19. PCPaul Copplestone8:16

       ... different tools. So it's a new, it would be a new model for these hyperscalers.

   20. SPSpeaker8:21

       Mm-hmm. Okay, so not a big worry. Uh, but well, uh, early on you did have that kind of success. Uh, did that take off right away? Like, when you re- how did you launch?

6. 8:33 – 9:59

   The "Open Source Firebase" Pivot

   1. PCPaul Copplestone8:33

      So yeah, our initial launch, well, our very initial launch was positioning ourselves as real-time Postgres.

   2. SPSpeaker8:40

      But Hacker News, I guess.

   3. PCPaul Copplestone8:42

      Yeah, so-

   4. SPSpeaker8:42

      Given that, given how much you were reading the-

   5. PCPaul Copplestone8:44

      Yeah, so what happened was, in my previous startup, um, I was developing the, the tech stack-

   6. SPSpeaker8:49

      Mm-hmm

   7. PCPaul Copplestone8:49

      ... knowing that I wanted to launch a DevTools company. I was using Postgres and Firebase.

   8. SPSpeaker8:56

      Mm-hmm.

   9. PCPaul Copplestone8:57

      And, uh, Firebase was great. We were using it for this chat application, but we hit some scaling limits, and we, um, had to find out how to solve these limits.

   10. SPSpeaker9:06

       Okay.

   11. PCPaul Copplestone9:07

       So I ended up building the real-time engine on top of Postgres, so one of the tools that I couldn't-

   12. SPSpeaker9:12

       Mm-hmm

   13. PCPaul Copplestone9:12

       ... find in the ecosystem. I put that on show Hacker News, and it started getting some trac-traction. And then I thought, "Oh, now's the time to build this database company."

   14. SPSpeaker9:23

       As a company.

   15. PCPaul Copplestone9:23

       And I thought-

   16. SPSpeaker9:23

       Okay

   17. PCPaul Copplestone9:23

       ... "Oh, well, the real-time engine, people will like that, so we'll call it real-time Postgres. That's our differentiator." And we kind of launched it. We applied for YC. We got into YC, but it wasn't really getting that much traction.

   18. SPSpeaker9:36

       Okay.

   19. PCPaul Copplestone9:37

       And then on one day, we changed the tagline from real-time Postgres to open source Firebase alternative. Um, and at that point, yeah, it kind of... Actually, someone else launched us. They put us on Hacker News, and it went straight to the top for a few days.

   20. SPSpeaker9:54

       Oh, okay.

   21. PCPaul Copplestone9:54

       And at that point, we knew we had something, but we were missing a lot of, a lot

7. 9:59 – 10:26

   Missing Features, Big Ambitions

   1. PCPaul Copplestone9:59

      of features.

   2. SPSpeaker9:59

      You were not yet Firebase-

   3. PCPaul Copplestone10:01

      No

   4. SPSpeaker10:02

      ... like-

   5. PCPaul Copplestone10:02

      No

   6. SPSpeaker10:02

      ... feature parity or anything.

   7. PCPaul Copplestone10:04

      No, exactly, and that was kind of aspirational. I mean, Firebase is a phenomenal, um, backend, and it's very comprehensive.

   8. SPSpeaker10:12

      Mm-hmm.

   9. PCPaul Copplestone10:12

      We knew we mostly wanted to be a database centric, and Firebase has much more than the database side of things. So we just decided on the few tools that we really wanted to get Postgres out there, and we started working towards those.

8. 10:26 – 12:18

   Winning Developers Over Firebase

   1. SPSpeaker10:26

      That's awesome. How would you convince, like, kind of like, uh, startups? Because I guess you started mostly from, like, bottom up with small companies first.

   2. PCPaul Copplestone10:35

      Yeah.

   3. SPSpeaker10:36

      How would you convince them to use you and not Firebase? I mean, Firebase was already backed by Google.

   4. PCPaul Copplestone10:41

      Yeah.

   5. SPSpeaker10:41

      Kind of like more reliable in the mind of people, I guess, than a new company just starting.

   6. PCPaul Copplestone10:48

      Yep, definitely.

   7. SPSpeaker10:49

      Yeah.

   8. PCPaul Copplestone10:49

      I think, um, the main things that we did... Well, actually, the, the not Google was [laughs] funnily enough one of the positionings that, that worked quite well, um, for, for us at least, or for the developer audience.

   9. SPSpeaker11:04

      Yeah.

   10. PCPaul Copplestone11:04

       Um, the other things that, the differentiators with, uh, Firebase, because really it's, Firebase is one of those tools that you, you, you both love and you might hate. Like, it's one of these phenomenally easy tools as a developer. So for us, we... The reason why you might dislike it is you might scale off it, or you might not like the pricing. So we solved those by saying, "Well, it's Postgres, and here's our pricing." You basically could not get bill shock from, from Supabase.

   11. SPSpeaker11:33

       Mm-hmm.

   12. PCPaul Copplestone11:34

       And then the other things that we said were that it's open source, which I know-

   13. SPSpeaker11:38

       Yes

   14. PCPaul Copplestone11:38

       ... James, the founder of Firebase, really wanted to do, but in the end they didn't pull it off after they moved to Google.

   15. SPSpeaker11:45

       Yeah, to Google.

   16. PCPaul Copplestone11:45

       Um, so being open source was something that developers really resonated with. And, um, the Postgres side of things meant that it was a relational database rather than a NoSQL database, and that ended up being a very important decision throughout the years. As the shift kind of moved back towards relations

   17. SPSpeaker12:02

       Oh, yeah. So Firebase was actually NoSQL-

   18. PCPaul Copplestone12:03

       Yeah

   19. SPSpeaker12:03

       ... backend. Okay. That, that's in- that's super interesting.

   20. PCPaul Copplestone12:06

       Yeah.

   21. SPSpeaker12:07

       Um, and so that open source aspect, that community you kind of built around yourself, was the, the, the first, uh, seed of that kind of crazy growth that came after?

9. 12:18 – 14:27

   Launch Waves & YC Batch Adoption

   1. PCPaul Copplestone12:18

      Yes, yes. And, you know, it comes in waves, right? So the first one was that launch.

   2. SPSpeaker12:24

      Yeah.

   3. PCPaul Copplestone12:24

      Then right before our demo day, we did another big launch for an auth product, and then after that we did another launch where we went from alpha to beta. And then-

   4. SPSpeaker12:31

      Yeah. And so everyone was using you, uh, in your batch?

   5. PCPaul Copplestone12:34

      In our batch, no one was using us. [laughs]

   6. SPSpeaker12:37

      [laughs] Awesome.

   7. PCPaul Copplestone12:37

      So we're like, well, okay, we were the first remote batch. Um, so-

   8. SPSpeaker12:41

      Yeah. Uh, so you, you didn't have the same connections to the-

   9. PCPaul Copplestone12:43

      Connection with everyone. Um, but also I wouldn't ex- have expected people. You have to build up some credibility when you're a database company, so.

   10. SPSpeaker12:51

       Even the YC founders are a little more-

   11. PCPaul Copplestone12:52

       Yeah

   12. SPSpeaker12:52

       ... conservative.

   13. PCPaul Copplestone12:53

       They're, they're, they're... A- and they should be, um-

   14. SPSpeaker12:56

       How about today?

   15. PCPaul Copplestone12:57

       So today, now we've got, yeah, what? 10 million developers using-

   16. SPSpeaker13:01

       Wow

   17. PCPaul Copplestone13:01

       ... Supabase, so it's a-

   18. SPSpeaker13:02

       And at, and at last, uh, YC companies use you?

   19. PCPaul Copplestone13:05

       Ah, so yeah, now every batch, I, I think on the last batch we had over 60% of the batch.

   20. SPSpeaker13:10

       60% of, uh-

   21. PCPaul Copplestone13:11

       Yeah

   22. SPSpeaker13:12

       ... a single batch. Okay. That's, uh, that's a show of confidence.

   23. PCPaul Copplestone13:15

       Yeah, that's much healthier.

   24. SPSpeaker13:17

       That's awesome. And, um, I think you speak a lot about, um, the experience of people. You focus on the kind of the developer experience using Supabase.

   25. PCPaul Copplestone13:25

       Yeah.

   26. SPSpeaker13:25

       Can you tell us a little more about what you're looking at?

   27. PCPaul Copplestone13:29

       Yeah. For developer experience can mean many things. Um, the way that we measured it at the start was-

   28. SPSpeaker13:35

       Mm-hmm

   29. PCPaul Copplestone13:35

       ... um, a metric time to value, and it was pretty basic. What we wanted to do was we took a, another Postgres service, um, an AWS one, and we literally measured how long it would take for a developer to sign up, launch a database, connect to it-

   30. SPSpeaker13:52

       Okay
10. 14:27 – 16:50

    How Agents Changed the Metric

    1. PCPaul Copplestone14:27

       Yeah.

    2. SPSpeaker14:27

       So until last year it was still eight minutes.

    3. PCPaul Copplestone14:28

       Yeah, something like that. [laughs] And then, um, yeah, and of course that changes. Now time to value is important, but as well, like with the agentic move, maybe you don't care about time. You might kick something off, walk away for 20 minutes while the agent is building and, um, come back, and at that point the value is not so much the time, but the ability to get to a res- result, uh, without further intervention.

    4. SPSpeaker14:53

       Mm-hmm.

    5. PCPaul Copplestone14:53

       So, um, you know, time to value I still think is, is, um, a key metric. But what that means is, is shifting.

    6. SPSpeaker15:01

       Hmm. Excellent. Um, as you were sharing, like it looks like the course of Supabase had like some, uh, big jump at some point of time.

    7. PCPaul Copplestone15:09

       Yep.

    8. SPSpeaker15:09

       It feels like the, the, the, the AI kind of disruption also impacted you like, uh, in a big way, uh, starting with what? Like kind of VAP coding applications using you?

    9. PCPaul Copplestone15:20

       Yeah, there was actually three chapters of AI in the database world-

    10. SPSpeaker15:24

        Yeah

    11. PCPaul Copplestone15:24

        ... at least. So the first chapter was when vector databases came out. I don't know if you-

    12. SPSpeaker15:29

        Oh, I remember that time

    13. PCPaul Copplestone15:30

        ... embeddings. Uh-

    14. SPSpeaker15:31

        Like PG-

    15. PCPaul Copplestone15:31

        Seems like the distant past

    16. SPSpeaker15:33

        ... PG Vector.

    17. PCPaul Copplestone15:33

        Yeah, yeah, yeah. Exactly.

    18. SPSpeaker15:34

        It feels like, uh, that, that's so old now. [laughs]

    19. PCPaul Copplestone15:36

        It's crazy. And of course now RAG is... Everyone saw RAG as this completely revolutionary thing-

    20. SPSpeaker15:42

        Yeah

    21. PCPaul Copplestone15:42

        ... and it has to happen.

    22. SPSpeaker15:43

        I remember how vector DB were the thing-

    23. PCPaul Copplestone15:46

        Yeah

    24. SPSpeaker15:46

        ... like to, to look at.

    25. PCPaul Copplestone15:47

        Yeah, yeah. And that was super interesting. Um-

    26. SPSpeaker15:50

        And, but did that actually impacted you, like people started using you be- because of PG Vector or?

    27. PCPaul Copplestone15:55

        Yes, yes. Um, so we were looking around for, um, how to do embeddings-

    28. SPSpeaker16:00

        Mm-hmm

    29. PCPaul Copplestone16:01

        ... and it, it's actually quite a cool open source story. Um, uh, I received an email from someone, and they said to me, "Oh, um, you know, I really want to use Supabase, but I need RAG."

    30. SPSpeaker16:13

        Mm-hmm.
11. 16:50 – 21:18

    Building the First Docs Chatbot

    1. PCPaul Copplestone16:50

       docs. And at the time no one had done this.

    2. SPSpeaker16:52

       When was that?

    3. PCPaul Copplestone16:52

       This was like start of '23, I think. What are we-

    4. SPSpeaker16:56

       So just-

    5. PCPaul Copplestone16:56

       What are we now?

    6. SPSpeaker16:57

       Just maybe-

    7. PCPaul Copplestone16:58

       '23, yeah, I think

    8. SPSpeaker16:58

       ... yeah, just after-

    9. PCPaul Copplestone16:59

       Or '24

    10. SPSpeaker16:59

        ... ChatGPT release. Like people were already used to chat in-

    11. PCPaul Copplestone17:03

        It was super early in the-

    12. SPSpeaker17:04

        Okay

    13. PCPaul Copplestone17:05

        ... in the like RAG stage, and then we, uh, uh, we were really scared. Yeah, so actually it was the start of, um, GPT. I, I don't know what model it was, but it was at the stage where everything was hallucinating.

    14. SPSpeaker17:18

        Mm-hmm.

    15. PCPaul Copplestone17:18

        And so we were very scared that it would make up stuff in our docs.

    16. SPSpeaker17:22

        And, and generate output.

    17. PCPaul Copplestone17:22

        Yeah, yeah. So we released it as, uh, do you remember Microsoft Clippy?

    18. SPSpeaker17:27

        Yeah, of course.

    19. PCPaul Copplestone17:27

        And Clippy would be completely unhelpful, and so we thought it would be funny to release it as a demo, but make the icon a Clippy-

    20. SPSpeaker17:34

        Oh, [laughs] okay

    21. PCPaul Copplestone17:34

        ... a Supabase Clippy icon so that if it hallucinated, at least you'd have a laugh. And then since then, of course, as we know, everyone's doing RAG on their docs. But it completely came from this, this contributor.

    22. SPSpeaker17:45

        It feels like, uh, yeah, it feels like PG Vector kind of like, uh, took the whole market of-

    23. PCPaul Copplestone17:50

        Yeah

    24. SPSpeaker17:50

        ... vector DB.

    25. PCPaul Copplestone17:51

        So, so we, we were, uh, well, the first to offer it, and of course Andrew had been developing it for many years.

    26. SPSpeaker17:59

        Mm-hmm.

    27. PCPaul Copplestone17:59

        Um, so it was, yeah, really credit to him for, for-

    28. SPSpeaker18:02

        Mm-hmm

    29. PCPaul Copplestone18:02

        ... leaning into it so hard. And then AWS as well adopted it, and from there it became the, the standard-

    30. SPSpeaker18:08

        The standard
12. 21:18 – 22:55

    Claude Code Changes Everything

    1. PCPaul Copplestone21:18

       year.

    2. SPSpeaker21:18

       Yeah, December release of new models. Yeah.

    3. PCPaul Copplestone21:19

       Which is Claude Code, and then that's where things felt fundamentally different because not only did our growth rate on the top of funnel increase more developers pouring in, but basically every other chart started going up, which was unusual. So our-

    4. SPSpeaker21:35

       What, what, what those are charts?

    5. PCPaul Copplestone21:37

       Activation, so getting to an active database. Um, then-

    6. SPSpeaker21:41

       So pe- people who would use Claude Code, uh, were doing way more than the people who would use Lovable.

    7. PCPaul Copplestone21:47

       Yes.

    8. SPSpeaker21:47

       And so they would activate faster-

    9. PCPaul Copplestone21:49

       Yeah

    10. SPSpeaker21:49

        ... and then use more.

    11. PCPaul Copplestone21:52

        They would get to a, let's take the average profile of someone using Claude Code. They would say, "Oh, I want to build this project."

    12. SPSpeaker22:00

        Mm-hmm.

    13. PCPaul Copplestone22:00

        And then within a day they would get to their conversion, uh, as in like it's a production database and I want to start paying for it, and then they would put it out and probably be more successful, and so the database grows even faster. So basically every curve was going up and to the right, which usually when your top of funnel is going up and to the right, a- as in it's accelerating off this base, it's because you're pouring marketing dollars into your-

    14. SPSpeaker22:25

        Yes

    15. PCPaul Copplestone22:26

        ... into your growth. We didn't, we don't really pay-

    16. SPSpeaker22:29

        You're not spending so much on marketing

    17. PCPaul Copplestone22:29

        ... we don't pay much for marketing at all. Uh, so the intent-- the quality of our audience is getting even higher, so it's, we're not diluting our, our audience at all. It's just high intent builders. And so because of this, Claude Code would get them more successful, to a successful stage faster. And, uh, yeah, just our, our revenue would accelerate, our conversion accelerate, everything accelerated.

13. 22:55 – 27:03

    60%+ of Databases Launched by Agents

    1. SPSpeaker22:55

       How many of your users are not humans anymore? [laughs] Like, like what's agents? What's humans?

    2. PCPaul Copplestone23:02

       Yeah. It's hard to say. I, I mean, of course they're all human kicking off their agents.

    3. SPSpeaker23:09

       Mm-hmm.

    4. PCPaul Copplestone23:09

       So, but from what we can actually measure, um, over 60%, uh, of databases are launched by an agent.

    5. SPSpeaker23:17

       Okay.

    6. PCPaul Copplestone23:18

       But more likely it's, it's even higher. We just can't tell.

    7. SPSpeaker23:20

       Yeah. I would, I would guess like 90 or...

    8. PCPaul Copplestone23:22

       Yeah. Yeah.

    9. SPSpeaker23:23

       That's fine.

    10. PCPaul Copplestone23:23

        Because they could-- If they're using Claude Code, they're just launching it from their CLI and we don't have-

    11. SPSpeaker23:28

        Oh, you don't know when they do that?

    12. PCPaul Copplestone23:29

        Yeah.

    13. SPSpeaker23:29

        Okay. So like you, you under count there-

    14. PCPaul Copplestone23:31

        Yeah

    15. SPSpeaker23:32

        ... by a big margin. What does that mean? Like in this world where the vast majority of like situation is like driven by an agent, by AI, what does that mean for developer experience? Like I guess you're not really targeting you- developers anymore.

    16. PCPaul Copplestone23:46

        I think the main thing is that the tooling will have to change so that it will work with agents.

    17. SPSpeaker23:52

        Mm-hmm.

    18. PCPaul Copplestone23:52

        That's obvious, and everyone is doing that. So what that means for Supabase is that if you're using MCP or CLI, these have to be first class-

    19. SPSpeaker24:01

        Yes

    20. PCPaul Copplestone24:01

        ... citizens. So in our earlier chapters, we were really focused on the dashboard experience of Supabase because a human would come in, and they'd play around, and they'd probably just activate-

    21. SPSpeaker24:11

        Do they still come in?

    22. PCPaul Copplestone24:12

        They, they do. Um, we still see our daily active users going up.

    23. SPSpeaker24:17

        Okay.

    24. PCPaul Copplestone24:17

        But not as rapidly. They, they can get to a successful-

    25. SPSpeaker24:21

        So the percentage of people who spend time on the dashboard is way lower than it was.

    26. PCPaul Copplestone24:25

        Much lower, yeah. Yeah, exactly. Then, um, so once the tooling changes to focus on CLIs, getting everything down into code-

    27. SPSpeaker24:34

        Mm-hmm

    28. PCPaul Copplestone24:34

        ... um, then you have to have really nice Git workflows.

    29. SPSpeaker24:38

        Mm-hmm.

    30. PCPaul Copplestone24:38

        So for us, we have a Supabase folder which should basically mimic what we have in the dashboard.
14. 27:03 – 31:17

    Scaling from 220 to 360 People

    1. SPSpeaker27:03

       company?

    2. PCPaul Copplestone27:04

       In the initial stages of Supabase, we had this really strict policy of hiring.

    3. SPSpeaker27:09

       Mm-hmm.

    4. PCPaul Copplestone27:09

       And our policy was you only hire when there is a hair on fire problem.

    5. SPSpeaker27:14

       Okay.

    6. PCPaul Copplestone27:15

       So something that a full-time employee could handle, and otherwise we just manage it. Then as we started growing, of course, the amount of users pouring in and the amount of support that we needed to give, and the amount of, uh, IP, IP addresses running out-

    7. SPSpeaker27:32

       Yeah

    8. PCPaul Copplestone27:32

       ... and all sorts of random things that you don't think about.

    9. SPSpeaker27:35

       Stuff like, yeah, you don't think about that-

    10. PCPaul Copplestone27:37

        No

    11. SPSpeaker27:37

        ... when you are small. [laughs]

    12. PCPaul Copplestone27:38

        No.

    13. SPSpeaker27:38

        For sure.

    14. PCPaul Copplestone27:38

        So what, what is like a planned roadmap ends up just being a lot of things coming in from the side and, and partnerships coming in that you can't say no to, and all sorts of stuff that, uh, happens. So the shift goes from, um, you pushing a lot to you as basically being a, a, um, a order maker to an order taker and, um, we have all these things coming in where we just didn't have enough hands on keyboard to, to solve them. And that of course means we've got to focus on reliability and, um, scaling up-

    15. SPSpeaker28:11

        Like data

    16. PCPaul Copplestone28:11

        ... the meta processes of the company, managers and hiring staff and everything else. So you go through that second stage of, uh, first you focus on your roadmap, then you focus on your business.

    17. SPSpeaker28:23

        How many people are you today?

    18. PCPaul Copplestone28:24

        Now we're about 360.

    19. SPSpeaker28:26

        Which is pretty small.

    20. PCPaul Copplestone28:29

        It's quite small, yeah.

    21. SPSpeaker28:30

        For the scale of the business.

    22. PCPaul Copplestone28:31

        But grown fast this year. I think we started the year at about 220, so we've had to-

    23. SPSpeaker28:35

        Okay

    24. PCPaul Copplestone28:36

        ... add a lot of people this year.

    25. SPSpeaker28:37

        It's all mostly to pretty much support the growth. The business is growing so fast.

    26. PCPaul Copplestone28:41

        Yes. All to... Yeah, so that's where we switched the-

    27. SPSpeaker28:44

        Uh, uh, can you give us a sense of how fast the business is growing?

    28. PCPaul Copplestone28:47

        I think maybe in terms of the number of developers, um, maybe we've in the past 12 months, at a guess, I think it would be 6 or 7X. Yeah.

    29. SPSpeaker28:59

        Okay.

    30. PCPaul Copplestone28:59

        So quite big.
15. 31:17 – 34:19

    AI Inside the Company

    1. PCPaul Copplestone31:17

       Of course, like a lot of, um, companies focus on like flashy terms like token maxing and things like that. Uh, we haven't done that.

    2. SPSpeaker31:26

       Mm-hmm.

    3. PCPaul Copplestone31:27

       But we also haven't had to. We- we've been very fortunate in one key thing. Supabase as a company is completely distributed. So-

    4. SPSpeaker31:35

       Yes

    5. PCPaul Copplestone31:35

       ... we have 360 people, and we're spread across 60 different countries, more than 60 countries. So this means everything that happens is written down on Slack, on Notion, and-

    6. SPSpeaker31:46

       And it's not like, uh, if it was a company that started doing that six months ago.

    7. PCPaul Copplestone31:49

       Yeah.

    8. SPSpeaker31:49

       You have done that for six years.

    9. PCPaul Copplestone31:51

       Yeah. And we've got the full history of Supabase written down all the way back to the start of the company. So if I want to find out what decision we made, uh, right at the start, I can ask Slap- SlackBot and find the answer.

    10. SPSpeaker32:03

        So you have the context. You have like, the complete context of the company-

    11. PCPaul Copplestone32:06

        Mm-hmm

    12. SPSpeaker32:07

        ... for like since inception-

    13. PCPaul Copplestone32:08

        Yeah

    14. SPSpeaker32:09

        ... available to AI.

    15. PCPaul Copplestone32:10

        Yes. Yes.

    16. SPSpeaker32:11

        And so when you use agents to help you in your job, they are like, this couldn't have more context.

    17. PCPaul Copplestone32:15

        Yeah, exactly. So new people coming in instead of-- or even, uh, they might not be new people.

    18. SPSpeaker32:20

        Mm-hmm.

    19. PCPaul Copplestone32:20

        They might just be someone who is, let's say, um, well, a CFO or an analyst who wants to understand something. Instead of coming to me and asking what the timelines are and, um, how do I explain this product to an investor, they can actually ask a bot, and the bot will explain it better than I can. So that's nice. And then from this point on, what I have traditionally thought of as just BizOps is-

    20. SPSpeaker32:46

        Mm-hmm

    21. PCPaul Copplestone32:46

        ... now AI, AI-enabled BizOps.

    22. SPSpeaker32:48

        Mm-hmm.

    23. PCPaul Copplestone32:49

        So we're looking at systems that we can automate end-to-end where, uh, we can introduce AI and automate the entire, um, process as long as, as it's an end-to-end process. So things like could be as simple as creating the AI, AI notes, but also generating them, finding the action items, and feeding it all into a company operating system, which hopefully at some stage we're thinking trying to move everything to Linear, including the non-engineering folks-

    24. SPSpeaker33:20

        Okay

    25. PCPaul Copplestone33:20

        ... so that we've got task management. And if we ever need to, to, um, you know, have bots solving things, of course, the engineers are the ones adopting these tools now.

    26. SPSpeaker33:29

        But it's there. Okay.

    27. PCPaul Copplestone33:30

        But as we develop it with engineering, then it will flow trickle out to non-engineering teams.

    28. SPSpeaker33:35

        So today you mostly will get a lot, uh, of productivity gain on the engineering side. But, but the next frontier is like every, every other function.

    29. PCPaul Copplestone33:44

        I think there's-- I wouldn't say productivity gain-

    30. SPSpeaker33:48

        Mm-hmm
16. 34:19 – 38:43

    True Gains Need No Human in the Loop

    1. PCPaul Copplestone34:19

       team.

    2. SPSpeaker34:19

       So you mean like the, the human become the bottleneck here because they have to approve everything?

    3. PCPaul Copplestone34:25

       Yeah. If you're-

    4. SPSpeaker34:26

       And you already have like some of these loops working without a human?

    5. PCPaul Copplestone34:30

       Yes, yes. So a, a good one was from our data engineering team. We use-- We, we now ingest everything into a data warehouse.

    6. SPSpeaker34:39

       Mm-hmm.

    7. PCPaul Copplestone34:41

       And, um, we use a tool called Hex, which is great.

    8. SPSpeaker34:43

       Mm-hmm.

    9. PCPaul Copplestone34:43

       And it's got an agent, and we now opened up Hex to our en-entire company. So we kind of format the data warehouse so that the agent will understand the context.

    10. SPSpeaker34:54

        Mm-hmm.

    11. PCPaul Copplestone34:55

        And this includes sales calls and things like that. So now if a, anyone in the company, a salesperson, CFO, a product person, wants to understand something, m- maybe a metric or our revenue growth-

    12. SPSpeaker35:09

        Mm-hmm

    13. PCPaul Copplestone35:09

        ... or, or even the history of a particular customer and what their needs are-

    14. SPSpeaker35:15

        Yes

    15. PCPaul Copplestone35:15

        ... they can go into this, and they can ask a question. They'll get a much more detailed answer than a, what a data engineer traditionally would produce. And this is good because we now can measure if a data engineer could, in a traditional sense, answer, say, 300 questions per month, um, we can see how many questions are being asked across the entire company. And it turns out that actually there are three times as many questions being asked, um, because of the way that they've engineered the data warehouse. So this is good. You can see-

    16. SPSpeaker35:49

        Yeah

    17. PCPaul Copplestone35:49

        ... that we're already getting some gains now. What we wanna see is that replicated out to other teams where you very much have people focused on the systems part of it, and if those systems can unlock higher productivity, then that is great. Yeah, I think we've talked in the past about our sales team is another really interesting one.

    18. SPSpeaker36:07

        Yes.

    19. PCPaul Copplestone36:07

        Yeah.

    20. SPSpeaker36:08

        Uh, where you can also-

    21. PCPaul Copplestone36:10

        Yeah, the sales team

    22. SPSpeaker36:11

        ... leverage AI.

    23. PCPaul Copplestone36:12

        So we traditionally operated on a PLG model. In fact-

    24. SPSpeaker36:16

        Yes

    25. PCPaul Copplestone36:16

        ... not even traditionally, it's all PLG. And then we thought, all right, we want to-

    26. SPSpeaker36:21

        Mm

    27. PCPaul Copplestone36:21

        ... introduce salespeople. But what you would typ- Typically do, as you would know-

    28. SPSpeaker36:26

        Yes, I know very well. [laughs]

    29. PCPaul Copplestone36:27

        ... introduce some salespeople who kind of flounder around a little bit trying to sell something. Um, but you know, we've got a bunch of data from our customers, and we also want them to be very helpful. So we can see that an account might have had a poor experience with the database, and so we can send, all right, 1,000 of these customers are growing very fast. They've hit some limits-

    30. SPSpeaker36:51

        Mm-hmm
17. 38:43 – 40:47

    $500M Round, Staying Grounded

    1. PCPaul Copplestone38:43

       Yeah. Well, it's a lot of money. [laughs]

    2. SPSpeaker38:45

       Yeah.

    3. PCPaul Copplestone38:46

       So, so, um, but of course, as you become bigger, you need a lot more operating cash to float you between two different periods, the period where you can launch a database and someone can build on top of it, and then they start paying you.

    4. SPSpeaker38:58

       Yes.

    5. PCPaul Copplestone38:58

       So we of course want to have that operating capital-

    6. SPSpeaker39:02

       Mm-hmm

    7. PCPaul Copplestone39:02

       ... to make sure that we can float between these two, and as long as we know the economics stack up, then it's very reasonable for us to continue doing that. So that's great. Uh, having the cash in the bank is basically to give us that operating capital. Then, um, from there, staying grounded is largely a factor of just remembering that the cash actually doesn't mean that you're successful. Uh, I think-

    8. SPSpeaker39:26

       Mm-hmm

    9. PCPaul Copplestone39:27

       ... actually YC drilled you into the, uh, drilled this into you, where the fundraise itself is just a meta process.

    10. SPSpeaker39:33

        It's a mean to an end. [laughs] It's not-

    11. PCPaul Copplestone39:35

        It's just a, yeah, exactly. And if having more cash was the thing that made you successful, then by all rights, Supabase would not exist. There are many of our competitors who have far more cash, huge companies, but still we exist because we did things that they don't.

    12. SPSpeaker39:51

        Mm-hmm.

    13. PCPaul Copplestone39:51

        And we should continue focusing on those things that we do very well as the way to continue to grow our, our stake in the market.

    14. SPSpeaker39:59

        Mm-hmm. Uh, how about the team itself? Like, uh, I mean, the being grounded as a founder is the first step. Uh, but when the team, uh, sees these numbers, it's easy for them to probably get a little lost.

    15. PCPaul Copplestone40:14

        A, a lot of it spills out from, from the early founders. I mean, our way of operating is very much like a YC model, and then you, you hire in a bunch of people who are also of the same mindset. We've got a bunch of YC founders that work there and a, a lot of ex-founders.

    16. SPSpeaker40:29

        Okay.

    17. PCPaul Copplestone40:30

        And this trickles out into, into the rest of the organization. And then, yeah, um, we try to treat our employees very, very well, but also be very transparent with them that, you know, we want to continue doing it in a way that we can look back and, and be

18. 40:47 – 41:42

    Advice for DevTool Founders Today

    1. PCPaul Copplestone40:47

       quite proud of.

    2. SPSpeaker40:47

       Um, to conclude, uh, what kind of advice could you give to other maybe DevTool founders, like who are starting today in this kind of, like unseen, [laughs] like unprecedented world they are navigating today?

    3. PCPaul Copplestone41:02

       The most generic advice that I can give, which they should already be doing, is, um, you have to be quite relentless and prolific. So you have to try many things, and when you find-

    4. SPSpeaker41:13

       Mm

    5. PCPaul Copplestone41:13

       ... the things that work, then you have to relentlessly push on that thing to see where, where it can continue to grow, and then you try many things around it as well. So as long as you have those two mindsets, then it basically translates to a, a grind mindset. Then, then, uh, you should hopefully, with a bit of luck, find some success.

    6. SPSpeaker41:33

       [laughs] And luck is a necessary component-

    7. PCPaul Copplestone41:35

       It's a very necessary component

    8. SPSpeaker41:37

       ... of any kind of success-

    9. PCPaul Copplestone41:37

       Yeah. Yeah

    10. SPSpeaker41:37

        ... uh, successful startup. Awesome. Thank you so much, Paul, for joining us today-

    11. PCPaul Copplestone41:40

        Thanks, Nicholas

    12. SPSpeaker41:40

        ... and sharing your story.

    13. PCPaul Copplestone41:41

        Thank you.

    14. SPSpeaker41:42

        Thank you.

Episode duration: 41:43