A recent project on HN declared Supabase dead: https://www.isthistechdead.com/supabase
While the funding is impressive, I haven’t come across too many people touting Supabase or using it in production.
While the funding is impressive, I haven’t come across too many people touting Supabase or using it in production.
It is good to get started and no doubt useful for simple CRUD apps. But once you want to start doing more complicated stuff, a lot of the RLS primitives become very hard to maintain and test, for example. You could say that that's Postgres's fault, but Supabase strongly pushes you in that direction.
The tooling, while looking quite polished, just felt pretty half baked along with docs (at least a year ago when we pulled the plug). Try to implement even a halfway complicated permissions scheme with it and RLS and you are in for a world of hurt and seemingly unmaintainable code.
So we ditched Supabase Auth for AuthJS, and are using vanilla postgres with Prisma. That's worked well for us. All the tooling is relatively mature, it's easy to write tests, etc.
Maybe if AI is writing some of the code, it might get easier, but for now, I'm avoiding Supabase like the plague until I see a project that's relatively complex that's actually easy to maintain.
My experience of supabase really demonstrates to me that the ideals of all of the postgres layer technologies - postrest, realtime via wal, jwt auth in the db -, just don't make for an easy experience. It all works (mostly) but I find it more annoying than useful and have to work around it more often than I'd like. I suppose I'm old school, but just building the things that one needs is often more robust and less work than trying to plug into what they've provided.
I really don't know what they're going to do with a series D. It seems they now _have_ to go for a high-value exit, but I really don't see which company would provide that exit.