“RLS is by far the best part about Supabase IMO but also the most dangerous. I remember the Lovable days when you would setup a Supabase project for it independently and now I think its just baked in.”
“I needed a backend for my first public project and many people on the internet recommended Supabase. Now 6 months later I think it's a great service. It makes my life as a solo developer probably as easy as it can get on the backend side.”
“I had only heard of Supabase before, and i knew they used Postgres (my favorite database) under the hood, so i migrated the project from Firebase to Supabase. So far i really like it!”
“It’s an excellent choice for backend engineers who prioritize speed and a centralized platform for managing essential functionalities such as authentication, storage, and databases. However, at scale, it will be expensive. In such cases, we opt to self-host, especially if our team possesses the necessary time and expertise.”
“Not off Lovable Cloud specifically, no — I've moved Supabase projects between instances, which is the same job once you can get at the database, and that "once" is the first thing I'd check. Before anything else: can you get a direct Postgres connection string out of Lovable Cloud?”
“I'm using Supabase for a browser card game with two to four players. The rules run in TypeScript on the acting player's client, and Supabase handles persistence and realtime updates.”
“Supabase is great for rapid prototyping but can be limiting for complex business logic. If you're finding the separation of concerns lacking, it might be worth considering a hybrid approach.”
“Another way to do this is Cloudflare Hyperdrive via a worker and you touch Hyperdrive via a worker binding, keeping Supabase out of public view. In our apps, we rub everything over rpc so there’s no routes exposed to an unauthed user anywhere.”
“Pretty easy to overcome the RLS issues in supabase if you just use Drizzle. Way easier for agents to write Typescript than it is for them to write and maintain RLS policies.”
“i’m always tempted to try new tools, but i usually end up going back to n8n, supabase, and codex anyway. once the stack works, switching really can create more work than it saves.”
“For me, Supabase gives strong Pg guarantees reachable by my LLM without needing additional subscriptions or config. The LLMs are wonderful for inference and triangulation, but horrid at deterministic data grinding at scale. The two go hand-in-hand.”
“With Supabase, the LLM workflow feels fractured. Too much context lives outside the code—in database settings, rules, and the web UI which just adds friction to every prompt.”
“a Supabase reset drops rows inside the database and there is nothing on disk to copy, so the honest answer there is `supabase db dump` before anything destructive, not a gate.”
“I'm not sure that Supabase was /ever/ genuinely relevant. It was a low-code solution for individuals (and sometimes whole teams) afraid of writing code.”
“Never had these issues with Supabase. It works very well. The only reason why ive seen people have issues is when they vibe code something and then it doesnt work.”
“The auth trigger failure mode is real, hit almost the same thing. But that's a Postgres/trigger problem more than a Supabase-specific one, you get the same vague failure surfacing with any BEFORE/AFTER trigger raising exceptions in a request path.”
“actually let me soften it a bit, if you want your site with authentication NOW and you don't mind to pay maybe supabase auth might be the right move for you, otherwise i see no scenario where it makes sense, and if you are bootstrapping projects it definitely does not make sense”
“It has always been the case, supabase or heroku are the things you used as a solo dev or as prototyping platform, otherwise you are asking for troubles and they cost a lot.”
“I use supabase bc I didn’t want the complexities of managing authentication. The only feature I use for supabase is that sign in with Google button, every other piece of information gets stored in database of my choosing.”
“supabase is used by major companies and is open source, you can pull it all and self host if you want, and having dealt with RLS and trying to run down the 500 error issue I see no compelling argument for why someone should make the call to abandon it for every project especially when you may need very clear delineation, meet HIPAA/ISO requirements, etc.”