I found a real production bug by reading a raw Postgres error — with zero coding background
Title: I found a real production bug by reading a raw Postgres error — with zero coding background
Three months ago I couldn't read a stack trace. Today I fixed a bug that was silently breaking check-ins for every user by tracing a foreign key error down to one line in a migration file.
Some context: I'm building Ember, an AI accountability app, completely solo, using AI coding tools to write basically everything. No CS degree, no prior engineering job, nothing. I direct the changes, review the diffs, and until pretty recently, I mostly just trusted that things worked if the screen looked right.
That stopped working the day a feature quietly broke, and nobody could tell me why.
Users could submit a check-in, but their score history never saved. No crash, no obvious error on screen, just silence where data should've been. I asked my AI tool to fix it twice. Twice it guessed wrong.
So I actually opened Vercel's logs myself. And there it was: insert or update on table "score_history" violates foreign key constraint "score_history_user_id_fkey"
I didn't know what that sentence meant six months ago. I do now. It meant one of my database tables was pointing at the wrong other table, a mismatch that had probably been sitting there since I built the feature, just never triggered until real usage hit it.
I fixed it with three lines of SQL. Took ten minutes once I actually knew what was wrong. The part that took three months was getting to the point where I could read the error at all.
What I didn't expect: reading logs turned out to be a completely learnable skill, not some invisible line separating "real engineers" from everyone else. The error message was just... telling me the answer, the whole time. I just had to stop being scared of it and actually read it.
Remember: If you're building solo with AI tools and something breaks in a way that feels impossible to understand, open the actual logs before you ask the AI to guess again. Half the time, the computer already told you exactly what's wrong.
Replies