Codex spins its wheels while Claude Opus shows far deeper understanding, user reports
iruletheworldmo · x · 2026-10-06
A developer running a complex project with AI for a long stretch reports that Codex kept spinning its wheels—creating more work rather than simplifying it—with most of its output being worthless "dead tokens." After letting Claude Opus review all the files, he found a significantly higher level of human-level understanding and far lower token usage since it doesn't waste effort. He plans to give Claude more scope on the project and says these are the only two models he'd trust with it so far.
More from coding & agent
- Open-source huashu-art-motion skill animates 35 art styles via coding agents — AlchainHust · 2026-10-06
- mysid: a native Rust MCP server giving Claude code graphs and clean file tools — techyphile7 · 2026-10-06
- Open-source plugin anonymizes names in Claude Code prompts to dodge account bans — vista8 · 2026-10-06
- LLM extraction silently dropped listings at chunk boundaries; 2K overlap and two-step dedup fixed it — InsideDebt6345 · 2026-10-06
- Her game hit 2M players and 100k+ concurrents in 4 days, yet people still call her a vibecoder — FrankFelixAI · 2026-10-06
- Creator shares 120-prompt workflow building an AI video game with Seedance 2.5, Godot and Magnific — techhalla · 2026-10-06