OpenAI Codex degraded for 1 h 2 min, shorter than its usual incident

What happened

OpenAI Codex is OpenAI's coding agent, available as an open-source terminal CLI, an IDE extension, a desktop app and a cloud agent that works on repositories. The status page lists the incident from Oct 21, 02:26 UTC to Oct 21, 03:28 UTC, with impact rated minor. The incident title also names ChatGPT and the APIs, but no individual components were listed, and the facts do not say what caused it.

How it compares

The typical Codex incident lasts 2 h 11 min, so this one ran 0.5x as long as usual. The longest incident in the past 12 months lasted 8 h 27 min. The count is the more notable figure: 7 incidents in the last 30 days, and the status page also shows 7 over the last 12 months.

Elsewhere at the same time

One other product had an incident open during this window: Lovable, with "Lovable backends stuck after restart", which ran 1 h 52 min. The facts do not link the two events.

What it means for you

• If you ran Codex jobs in the early hours of Oct 21 UTC, such as CI tasks or cloud agent runs, check that they finished.
• Codex had 7 incidents in the last 30 days, so a fallback coding agent for critical jobs is worth having.
• Subscribe to alerts on AgentGid or the official status page so you hear about the next incident when it opens, not after a run fails.