April 20, 2026
AI naps, the web snaps
ChatGPT and Codex Down
ChatGPT and Codex Down — users panic, blame “internet gremlins” and the web’s phonebook
TLDR: OpenAI’s ChatGPT, Codex, and API had a partial outage that’s now fixed, after a few hours of “investigating” and “monitoring.” Commenters split between blaming a wider internet wobble, joking it was DNS (the web’s “phonebook”), and sharing the status link—proof we’re all very dependent on our robot helpers.
For a few tense hours, people trying to use ChatGPT, Codex, and OpenAI’s API hit a wall — and the internet did what it does best: freaked out, theorized wildly, and posted links. OpenAI’s own status page logged a partial outage from mid-afternoon, flipping through Investigating to Monitoring before finally landing on Resolved by early evening. Translation: bots took a surprise nap, engineers poked them awake.
But the comments? Absolute gold. One user warned it wasn’t just OpenAI, noting Reddit also sputtered and that outage trackers lit up: “looks like more might be down than just openai,” cried the crowd, imagining a mini-internet wobble. In swooped the classic tech blame game: “I’m blaming DNS, unless proven otherwise,” joked another, invoking the running gag that the web’s “phonebook” (the system that turns names like openai.com into addresses) gets blamed for everything. And of course, the forum’s designated hall monitor dropped the must-click: status.openai.com.
Between refresh-button warriors and meme-lords chanting “internet gremlins,” the vibe swung from mild panic to gallows humor. Developers grumbled about frozen workflows, while others quipped that the outage was “forced time off.” OpenAI says all systems are back, but the bigger takeaway is clear: when AI gets the hiccups, the whole internet reaches for popcorn — and a scapegoat named DNS.
Key Points
- •OpenAI reported a partial outage affecting ChatGPT, Codex, and the API Platform on Apr 20, 2026.
- •Initial status was “Investigating” starting at 2:35 PM, with multiple follow-up investigation updates.
- •Mitigation was applied and status moved to “Monitoring” at 4:48 PM, with further monitoring at 5:43 PM and 6:17 PM.
- •The incident was marked “Resolved” at 6:48 PM, with all impacted services fully recovered.
- •Availability metrics are reported in aggregate; individual customer availability may vary by tier, model, and API features.