I gave GPT-5 a test run this afternoon using the free tokens in Cursor and thought I’d share some feedback.
I built out a version history feature for my prompt manager in about 5 minutes, zero bugs. Took a couple of iterations to fix minor CSS issues, but that was it. Pretty wild how quickly and accurately it all came together.
I originally built the first version of my prompt manager using Bolt, and over the past few months I’ve been tweaking it with Cursor. The stack is dead simple: Vue + Tailwind + Vite, auto-deployed to Netlify on every push to main in GitHub. Supabase handles the database (not the most secure setup in the world, but for a non-sensitive personal project, I really don’t care).
What GPT-5 handled for me:
- Full UI update (side-by-side diffs + restore)
- SQL for versioning logic
- Layout refresh to make prompts easier to read/edit
I know folks have had mixed experiences depending on the model, but so far I’m genuinely impressed!