Vibe coding + SQL

TLDR: I had to prompt to generate SQL queries while working with lovable.dev, and the results were interesting.

Background

Lovable.dev was promoted in a LinkedIn post, and its post on vibe coding promised Full-scale apps from simple descriptions.

It seems to target non-developers with posts like these, so I thought: if a non-technical person can do that, how much further can a technical person go?

Testing the hypothesis

Instead of creating an app from prompts I decided to remix an existing one to save time. I chose a note-taking app. The UI looks clean and more importantly, it works.

I tried a few prompts to add fields including an Author field, and it worked. Note-taking app from lovable.dev

It's frontend only and uses the React library; I tried editing it and found editing is disallowed in the Free plan.

The frontend app uses React and is structured

Lovable's Free plan also limits me to 5 queries per day. At this point I'm tempted to download the code and edit it myself. I'm a full-stack dev, I can do lots of things...

Hold on; lovable's promise was one can get a full-stack app, complete with databases, by merely prompting. It promised: Full-scale apps from simple descriptions.

I should at least test out its database connection through Supabase.

Database integration

I setup a Supabase account and successfully connected it to lovable. Given I chose a community project in lovable I expected to see ready-made SQL statements to create the tables. However there weren't, and I was asked to prompt to create the tables.

Hmm, there's a first time for everything.

I've written SQL CREATE statements before and never prompted for them. Here's what I did:

1. write the SQL CREATE statements and test it locally on Postgres
2. write a prompt to describe the statement
3. put the prompt into lovable and verify the resulting SQL was correct

A non-technical person can prompt to create the CREATE statements and get them right; this is not the hard part.

I then wrote a SQL query to list notes with tags. Since notes and tags is a many-to-many relationship I had to JOIN various tables and use STRING_AGG() to combine multiple rows. I wrote the query, tested it in SQL, and gave up with prompting for it via Lovable's LLM.

Prompting is too much work in addition to dev work. At this point I can recreate the tables + frontend locally. Prompting adds more time from ideating + testing + complexity.

I'm sure a non-technical person can figure out the query by prompting eventually, however SQL and prompting are separate skills, and it's a lot to learn both.

Does it match the hype?

If the hype is creating the app of your dreams with a few lines of English, then we're not there yet.

Lovable does give people a set of building blocks, and saves time choosing a technology stack.

This text-to-app category includes many players, including Google. The bigger players have higher usage limits while smaller providers like Lovable provide a focused feature set (eg. connection to Supabase).

Every platform is trying to be sticky and provides a custom prompt manual, including Lovable. Lovable's prompt manual is specific to app-building.

As a technical person I'm watching this space to save time on my development flow. So far prompting has not saved me time, but who knows, it might in the future.