I've been using Codex for about a week while attempting to code a web app from scratch. It generates a lot of code, but there are frequent omissions that require me to go back and fix things over and over. After three frustrating days, I decided to give Google's AI Studio a shot. Despite my skepticism toward Google due to past experiences with their AI and Google Home not working as expected, I'm surprisingly impressed with how well it builds and previews code. Has anyone else experienced better results with Codex, or does Google AI Studio simply outperform it right now?
1 Answer
Which version of Codex are you using? If it's the Codex CLI or Codex Agent from the ChatGPT app, the capabilities can differ. Generally, using models like o3 for architecture and o4-mini or 4.1 for implementation is great for coding. Plus, Gemini-2.5-pro is fantastic for frontend tasks, boasting better design abilities compared to 4.1. You might want to explore that for your project.
I was contemplating that! Maybe use Gemini for the frontend first in MOCK mode, then have Codex handle the backend.