My guy, it's a tradeoff of autonomy vs thoroughness. You might not enjoy using the codex models, but to say they're way worse than claude is an error.