Chapter 4 - taking a beating… a big one, from Lovable + Edge Functions. ~15 hours later, finally making progress. Doing back end in Lovable is waaay harder than front end.
First attempt: writing a long design doc (10 pages) -> It was good for ~absolutely nothing~ not much;
Drawing and sending the flow (it had worked before for the hooks) didn’t work either;
I gave up and opened the code, changed a few things, and then thought I should persist with the project;
I learned about the edge function logs (it took me a while to understand), dealt with CORS errors, and now it got stuck in loops with the errors I present from the Supabase logs.
Then it clicked: copy the code and the errors and throw them into ChatGPT o3-mini. It worked! I got unstuck.
Now at least they’re new errors. It seems to be moving again.
New lesson: it would probably be faster with a dev + Cursor, but the focus here is on the autonomy of a product person with a (no-code-friendly) IDE.
My guide so far:
- PRD + reference images help a lot (emphasis on the images)
- Refactor when the dev team (aka Lovable) asks; it helps a lot to keep the code organized ~and breaks a bunch of things too~
- Having trouble explaining the flow in text? Draw it. It works… partly
- Ask it to “log” everything on both the front end and the back end
- Use some “reasoning” model when you get stuck, to get out of the loop your “IDE for PMs/Designers” falls into.
- Ask the IDE to explain what the code in file X does; it helps you understand how to navigate the files.
- Ask it to clean up unnecessary code whenever there’s a big change
That’s it, losing this one so far… But it’ll get done. I didn’t think it would be easy, just as no-code wasn’t 10 years ago. :)
Originally written in Portuguese and published on LinkedIn on April 14, 2025. I translated it into English with the help of AI (Claude): it let me bring my LinkedIn posts to people who don’t read Portuguese without rewriting each one from scratch. I reviewed the translation to keep my original meaning and tone.