Blog

AI coding help — questions so production actually holds

Generic ai coding help builds pretty previews. These prompts force auth, env vars, payments, and database policies before you burn credits on CSS.

v0 supabase builds the UI — you still need the backend

v0 supabase generated a polished frontend. Auth, row policies, and env vars on your host still decide whether data persists after deploy.

Replit Agent changed too much and the database is gone

You asked replit agent for one UI tweak. It refactored half the Repl, ran migrations, and left test rows where your users used to be.

Replit deploy, then dead — no hosting after git push

You clicked replit deploy and got a URL. It loads empty, 502s, or never updates after git push — while the Repl preview still runs.

Lovable GitHub export — code out, then what?

You connected lovable github and pushed a repo. The code left the editor — hosting, SSL, env vars, and auth URLs did not follow automatically.

Lovable custom domain with no SSL — launch blocked

You attached a lovable custom domain and the browser warns Not secure, shows certificate errors, or never resolves at all while the default Lovable URL still works.

Lovable not working after publish — white screen from preview

Preview looked finished. After publish, lovable not working means a white screen, frozen spinner, or routes that load nothing while the editor still works.

replit credits vanished in one afternoon with the Agent

You opened Replit Agent after lunch and hit zero replit credits before dinner. Refactors, redeploys, and fix-the-fix prompts burned the monthly pool while production settings stayed wrong.

cursor agent mode rewrote files you never asked to touch

cursor agent mode autonomously edits across your repo. It ignored your rules, refactored Stripe routes, and left Composer-sized tasks looking cheap until usage dropped on the third unsolicited…