Codex nearly unusable: 51-task goal runs 46 hours, only half done, says user
manuelkoelman · x · 2026-10-08
Founder Manuel Koelman reports Codex has become "nearly unusable": a goal task scoped into 51 small chunks has run for over 1 day 22 hours with only 24 tasks finished. Even enabling fast mode five hours ago barely helped. A concrete, data-backed complaint about OpenAI's coding agent throughput and reliability for long-running tasks, relevant to any developer relying on Codex for extended autonomous work.
More from coding & agent
- Steve Yegge: Rex Alpha Makes 20+ Claude Code Sessions Feel Local From a Hotel Laptop — Steve_Yegge · 2026-10-09
- Agent harnesses are crutches; least-privilege action authorization is what matters — andreisavu · 2026-10-09
- Running Claude Code in YOLO Mode Without Getting Burned: A Practical Guide — Arindam_1729 · 2026-10-09
- LangChain open-sources Restock: an agent that buys and pays in Slack via Stripe Link — LangChain · 2026-10-09
- Workers Refuse to Write Markdown Files for Agents, Seeing Knowledge Extraction — mattbeane · 2026-10-08
- Fallout: New York runs in your browser, built with Claude Opus 5.5, zero texture or sound files — chrisfirst · 2026-10-08