GPT-5.6 models in codex are great at not writing like this. Cheaper, huge usage limits, readable output, and quality code. It's the better system now.