Update: OpenAI's 4-day outage streak is over; systems restored
Status is green again after four disruptions in four days. What broke, why the pattern still matters, and whether to unwind the fallbacks you just wired.
Copy markdown
Green across the board
OpenAI's status page now reports no known issues: API, ChatGPT and Codex are all operational after Saturday's failure returned 503 'circuit_open' errors that blocked requests before they reached the servers.
Four outages in four days
Saturday's hit followed disruptions on July 23 and 24. OpenAI's own status log shows roughly 166 incidents over nine months — about 18 a month — with ChatGPT posting the weakest 90-day uptime of the three services.
Why it stings more this time
With autonomous agents now live for around 10M weekly users, a mid-task API failure cascades through connected tools instead of just greying out a chat box. OpenAI moved from 'investigating' to 'monitoring' within an hour but never disclosed a cause.
What to do: keep the failover wired
You can safely route critical paths back to OpenAI, but four hits in four days is the case for leaving a second provider one config flag away rather than ripping out the fallback you added this week.