AMA w/ Supabase CEO

 @rrhoover thanks for the kind words - and thanks for everything you've built. Having a platform like PH is life-changing for companies finding PMF

> biggest surprise building the co so far

being able to build a successful remote/distributed company. When I started I thought I'd need to move to SF. We got into YC and were gearing up for the move but then COVID happened and we were forced to do it remotely. Since then we leaned-in to the distributed culture.

Many people told us we wouldn't be successful if we stayed remote/distributed. For many companies I think that's true, but for supabase it has been the opposite - we hire open source maintainers from anywhere in the world. These are the type of people we love: they are passionate about the technology, low-ego, and we don't need to do technical interviews to see if they are any good.

YC is still a big part of our DNA, so i'm in SF often, but it has been great being able to meet the community all around the world while I move around

@rrhoover @copple This observation hits close to home, as my last startup was @Tandem (lets remote teams work more like in-person teams). Despite working on this problem for years, I came away with the conclusion that in-person is better, and hard to replicate. A lot of other virtual office founders have had a similar conclusion.

There's conventional wisdom for remote companies: good documentation, intentional communication practices, regular team in-person gatherings. Is there anything unconventional that you've found is really important for Supabase? Or was it a compilation of small iterations and improvements?

@rrhoover @rajiv_ayyangar the #1 thing is to hire people who love distributed work. Not "remote" work, but "distributed" work. At the start we had a few mis-hires: people who had worked remotely, but what that meant was their whole company was in the same timezone and they'd all be online at the same time

At supabase, especially at the start, you could go a whole week and only have one 30 minute meeting. Not everyone likes that, but some people thrive on it. Open source maintainers are typically the profile who thrive in that environment.

We became very good at hiring people who would love working at supabase. Everything else you say is true too - good docs, written culture, etc

@rajiv_ayyangar we're targeting March 31 to April 4 for the next Launch Week

how can builders get involved

We run global meetups during launch week. Last time we had meetups in ~80 cities around the world. If you're keen to host a meetup just send us a DM on twitter: https://x.com/supabase

@rajiv_ayyangar 

do you do classic usability / ux interviews?

not really - at least not at the start. Most of it started with intuition - we knew what good looked like from experience. From there, developers are very good at giving feedback (sometimes in a very blunt way). We just listen to the feedback and iterate

@rohan_arun1 did you see this one already?

https://github.com/supabase-community/firebase-to-supabase

an API for auth

we have that too: https://supabase.com/docs/reference/api/v1-get-auth-service-config

Let us know if anything is missing

@copple yep but if you check my video my customers are completely non-technical so I would lose them at step 1 and I was looking for a 'magical' solution that would use auth to do it entirely in the browser like we can generate those impossible projects now that most devs wouldn't waste time on a year ago

The API doesn't allow setting up Auth from scratch last time I tried this and looking at those endpoint it still doesn't seem possible correct? Let's say I wanted to setup google auth for a vercel project entirely programmatically could we do that now?

@steveb we're only scratching the surface of Postgres - there is a lot more to do to make it easier and more scalable. We have a new Postgres storage engine coming out this year that is up to 5x faster than the current engine.

What risks and opportunities are there with the rise of AI?

we're iterating fast with apps like Bolt/Lovable/v0. We're fortunate that we offer a full backend - it's an important to have tightly integrated primitives so that there is less code. The less code required for building apps, the less likely the LLM is to make mistakes

We've already doubled in traffic since Bolt & Lovable launched ~2 months ago (something I didn't realise was possible). This is because it's a lot easier to build now - previously Supabase was very targeted at developers. Now it's accessible to a huge number of builders, even if you don't have a developer background. We're seeing a lot of new users "learning how to build" with Supabase - it's something i'm happy about since we are teaching them to use industry-standard tools like Postgres

@copple wild growth, that's amazing. Do you see the growth of things like Bolt/Lovable/v0 as a risk at all? Couldn't a larger company make "we lose money, but win market" deals with them to block you out?

@steveb strangely, these platforms aren't directly contributing to all of the growth - there is a halo effect that happens where more people talk about supabase as a result of these platforms and so more people use it directly. Losing a platform isn't a key business risk, but the platforms have been a huge boon to our online presence

we love being a backend for these platforms and developing tooling for them. There is a lot of detailed knowledge that goes overlooked: programmatic APIs, MCP servers, tailored prompts, etc. A competitor might undercut us, but if their platform is harder (for end users and LLMs) to use than supabase it's a net-loss for the AI Builder. We've spent 5 years building the best tooling possible, so I know it's not easy to pull off

@gabe firebase is an amazing tool - we take a lot of inspiration from it

If you want to us AI to develop with supabase it's worth checking out Bolt/Lovable/v0/etc - all of them have built very nice integrations with supabase. We also have our own built-in assistant in the dashboard (hit cmd+i)

fwiw, Supabase is a relational database (vs NoSQL). I see this as a big benefit for AI since LLMs have a tendency to hallucinate. In a database context this means they will insert all sorts of data into a database without constraints. A relational database provides the AI with a good feedback mechanism when they try to insert erroneous data