Lucian Ghinda · 2026-08-21 · notable
Lucian Ghinda — a week of reaching for Codex instead of Claude Code
Lucian Ghinda spent a week using OpenAI's Codex more than Claude Code. He reports that Codex writes simpler Ruby with fewer comments, while Claude Code anticipates edge cases and fits his CLI-based Jira workflow better.
A week-long, side-by-side account of Codex and Claude Code on the same real Ruby on Rails codebase.
What is it?
A week of using Codex more than Claude is a hands-on report from Ruby developer Lucian Ghinda, who moved his daily coding work to OpenAI's Codex and wrote down where it beat Claude Code and where it did not. Ghinda says Codex produced Ruby and Rails changes with fewer comments, which he preferred, and reached for simpler, more contained solutions.
How does it work?
To close the starting feature gap, Ghinda ported his Claude Code skills across to Codex, then ran several small, focused Codex sessions rather than one long sprawling one. He reports that Codex made fast first changes but needed heavy testing and review afterwards, so he saw no net time saving. He also hit nested branch structures Codex created and had to spell out rebasing.
Why does it matter?
Benchmarks rank coding agents; Ghinda's post instead describes how the two feel to work with over a full week on one codebase, and it reached the Hacker News front page with 160 points. His conclusion is about temperament: Claude Code anticipates what he will need next and builds more abstraction, while Codex follows the instruction conservatively and does not overextend.
Who is it for?
developers choosing between coding agents