Im developing enterprise software, making it robust to work in the real messy world and i must say, Codex is by far the most enjoyment i’ve had with coding agents.
claude just… does… it explains you and gives you a bunch of text at the end which i find fine but my eyes start to hurt after hours of reading paragraphs, code is fine but the “speed” is what annoys me a lil bit.
Codex tho… man what a ride. its like having my ruthless senior billionaire mentor stress testing me and making me understand everything as we build, i ask it questions and it responds so fast i can actually talk to the model and ask all the stupid questions (no question is stupid).
as a software dev i get how claude could be better to just get it done (and it does it well) but codex does the same but cleaner, maybe slower if you ask it questions and fool around trying to understand some things, overall this has been a super exciting ride and i hope the openAI team keep making their model better and aligned with its core principles, which i really love.