It’s like next.js but for backend.

Personally not for me. These are very productive once you learn the shape they expect.

But the wheel needs to be reinvented on how to do everything.

I ran have straight forward verbose code with no hidden control flow

I let the LLM tokens flow. Usually it ends up saving me innovation tokens.

To be honest, the use of AI in your marketing doesn't bother me. But paired with the fact that you have a newish HN account with no comments ever, and no comments on your own "Show HN" in response to others, doesn't put off strong trust signals.

Hey! I created a new account on HN just to have an account specifically for Typebase and I'm sorry I dind't answer more quickly the post got traction a during the weekend a couple of days after I posted. I'm answering everything right now!

The pronunciation of "Vercel" near the end of the "Typebase in 100s" video was VER-cel, with the emphasis on the first syllable, instead of ver-CEL with the emphasis on the second syllable.

Was the video generated by AI?

The audio is actually from a guy I hired on Fiverr but you are right that the pronunciation sounds kind of weird on Vercel.

Duh

I have deployed and maintained plenty of diff setups on GCP, AWS, colo, and I have also used AI for coding, my concern is do we really trust AI to do prod deploy and maintenance of the backend yet? Not to mention what are the limitations here, someone already mentioned queues? But for a straightforward site with limited complexities, it might be cool to try out.

Hey! Even tho I use AI to code faster I still review every line of code and I test every new feature. About the limitations, I'm a single developer working after work trying to add as many features as fast as I can so some of them may take more time to be ready. The good thing about Typebase is that under the hood is a normal TypeScript server so you can add any feature you want even if Typebase doesnt implemented natively!

I love the (well-disclosed) fake testimonials. Hilarious!

Thanks! I'm glad you liked them!

> “I’ve told four separate therapists about Typebase. Two stopped taking me as a client. The other two are now shipping an app with it.”

So good. I laughed out loud at this.

Scratching my head a little bit here, I'm not sure how this differs from NextJS (or Vite) with a layer of drizzle & better-auth.

Not that that layer isn't important. We ended up building something like this at Ingram, nextkit: https://github.com/ingram-technologies/nextkit - but we distribute it as a set of micro libraries that work well together rather than a unified framework. (Auth is nk-auth, built on top of betterauth, and sql is nk-db, built on top of drizzle)

You got to pretty much the same conclusions we did though regarding the "why".

I’ve been curious about trying the Supabase type of RLS model for an app but haven’t yet. I like the idea of the db defined api like postgraphile and hasura but had trouble with testing and ergonomics. This looks pretty nice though, it’s cool that you’re thinking about ergonomics for working with coding agents. I feel like a hard thing with that rn is that everyone is kind of figuring out different workflows for AI coding right now so it’s hard to optimize DX. Can you talk a little about your personal agent workflow and how this fits it?

I've only had experience in one full-stack project with supabase RLS so experience is limited. However it seemed like a mess. Seems much more human-maintainable to have a classic spring/golang backend. For sure in some projects, esp mobile apps, it makes more sense, and fable-level models might also keep it from being full spaghetti.

Also Supabase at that time wasn't reliable and it's not really made for self-hosting. Felt almost as bad as Github stability these days.

Still, based on this limited experience, I've been avoiding RLS and supabase.

Sounds very interesting. I have to agree I didn't like RLS as well. I mean, it's useful but it seems too many knobs to turn...

Small thing but it's "shoot yourself in the foot," not "shot" — sorry, can't help it. That said, yeah, RLS policies silently failing open because of a typo'd condition is a real footgun. Wish Postgres warned you.

Yeah I had the same problem, I think is super easy to shot yourself by using RLS

I was about the ask a question but you just answered how it is built. I guess I will try it out, it sounds really good.

I love your approach, fake testemonials and all.

Thanks!

Fake reviews are a hard no-no imo. Really weird addition, but I suppose it wasn't you who got the idea.

Looks an awful lot like a orpc, just built on top of drizzle instead of raw queries. I'm sure some people want this, but it seems like gross duct tape that will bite you in three years

Hey! I actually added the fake reviews myself, I wanted to give some personality to the landing.

> Looks an awful lot like a orpc, just built on top of drizzle instead of raw queries

Typebase is built on top of oRPC, the idea with Typebase is that you can build everything on super simple way and Typebase will take care of the rest.

I’m very surprised to see others commenting here that they like the fake reviews. Claude loves adding that shit, but you’re allowed to modify and remove the junk it produces.

> * None of these people exist. We checked. Twice. Legal is chill

Maybe I’m just the old man yelling at the clouds, but can we not have some integrity and actually put the truth out there? There is nothing wrong with using AI, but we’re still responsible for our work.

Reminds me of the "customer testimonials" era of 90s shareware, half of them written by the author under a fake name. Nobody's legal team ever got involved because nobody sued over a lie that small. This one just admits it upfront.

I don't know, I have plenty of gray hair, but I still love a little tongue in cheek humor. I still remember the first cold email I got asking for 12 minutes of my time - I replied. Now if the disclaimer was buried at the bottom of the page and this product was asking for money, I would agree there is more of a trust issue here.

Very cool!

Note that you can run Convex on top of your own Postgres, and we do this for our internal apps, with very nice performance (plenty for our internal use cases).

Yeah, that's the part people gloss over. Postgres underneath doesn't save you from rewriting queries in Convex's query language if you ever want to leave. We looked at it for a client project and the lock-in math didn't change much either way.

Thanks! But even if you run Convex on your own Postgres you still need to follow the Convex DB model right?

very cool. though the deploy wasnt clear if i can just run migration on some pg url. and it wasnt clear if i can bind some middleware into my existing app server (like express)

You can use your own DB by saving the URL on DATABASE_URL inside .env. For now migrations are not supported but they are coming on the next version. Another feature that is on the pipeline is being able to generate the server so you can bind it to your current server

What, in your opinion, is this not a good fit for?

I think on of the best things about Typebase is that it allows you to have typed actions you can call from your backend as if it where functions of your front end, for that reason I dont think Typebase is a great tool when you have apps on different languages apart from TypeScript

Enterprise software, stuff that you expect to scale exactly because it has queues and distributed computing.

Looks useful, congrats

Thanks!

This looks great. Question specifically on the Cloudflare Workers target: when you deploy there, does Typebase use D1 for the DB, or does the Postgres/Neon backend stay the same regardless of which server target you pick (Vercel/Workers/Deno) - i.e. is the DB layer decoupled from the server runtime, or does choosing Workers also mean choosing D1? I ask because I've shipped a small Workers+D1 service and the main friction I've hit isn't the server code, it's Workers' constraints around D1 (no long-running connections, HTTP-based driver quirks) - curious whether Typebase abstracts that away or whether you still feel it at the DB layer when targeting Workers specifically.

I see no code there to automatically pick D1 for the DB when you deploy to Cloudflare.

On Cloudflare, it should use the `@neondatabase/serverless` driver configured here https://github.com/typebase-io/monorepo/blob/fa3af0b71126caa...

Hey! For now we only support Neon but the idea is too add more DB providers soon. I wanted to build a "base layer" to show that Typebase is versatile enough, I'll add D1 to the priorities

Calling it decoupled seems generous if there's only one working DB provider. "Deploy anywhere" doesn't mean much when the anywhere still needs a Postgres connection string. Feels like the runtime target list outpaced the actual DB abstraction.

What happens on schema migrations once you've got real data in prod? Single-folder codegen is great greenfield, but every "write your backend as one artifact" tool hits the wall when a column rename needs a backfill script that doesn't fit the model.