Thibault Sottiaux, engineering lead on OpenAI's Codex team, says the coding tool maintains 99.98% uptime and sleeps only eight minutes per month
AI Judge changed title after evaluation, original title: "Thibault Sottiaux, engineering lead on OpenAI's Codex team, posted that Codex recorded 99.98 percent uptime limiting total downtime to roughly eight minutes per month"
Replies called the downtime unintentional and compared reliability to Stripe
Many users praise Codex's 99.98% uptime as incredible work by a dedicated team and better than Claude, while others complain about restrictive usage limits, slow resets, and poor frontend performance.
No Digg Deeper questions have been answered for this story yet.
Most Activity
Midnight thoughts
With 99.98% uptime, Codex only sleeps 8 minutes per month.
@thsottiaux When I look at codex of time I think about stripe of time and how stripe has a immediate effect on the GDP of the world if it's down and I think the same must be true for Codex maybe not the same immediacy but definitely the same effect
With 99.98% uptime, Codex only sleeps 8 minutes per month.
@thsottiaux unintentional sleeps
With 99.98% uptime, Codex only sleeps 8 minutes per month.
@thsottiaux Hey Tibo which department is the Prompting Cookbook ? I went through the Cookbook and tried to find issues and there were barely any. You run a tight ship. Accept my PR though <3 https://github.com/openai/openai-cookbook/pull/2710/changes/BASE..ae6983376d98fa5759389f2b1572ac7d0657d1f8#top
@thsottiaux That’s great, but uptime should be 99.99 % not 98, there is still the scope for improvement
@thsottiaux Happily Codexing
@thsottiaux 0.02% downtime × 5m users = ~43.2m user-minutes per month.
~82 years of collective waiting every month.
MIDNIGHT. THOUGHTS.
@thsottiaux A reset will boost codex to 99.999% uptime
@kennedyumege @thsottiaux My biggest concern is that as a pro subscriber I need to know ahead of time if we are getting less usage than previously. You can’t manage if you don’t know anything. 30% on Monday for minimal work is scary tight. I should need to try kinda hard to reach limit on Pro.
@kennedyumege @thsottiaux Same, it seems to be draining soooooo much quicker than it has been. Like very quickly. I don’t think I have ever been under 60% before at any point. Now, I’m at 30% on Monday and doesn’t reset until the 23rd!!!!!
Like, wtf?
@thsottiaux Bro, compaction essentially broken right now. Dunno about everyone but for me it doesn't work most of the time. Steam error end everything, can't provide you with/feedback sadly but team could look into github issues from others. Tried to set compaction earlier but still issues.
@thsottiaux This getting fixed? Had way more than 14% left last night. And can't even use codex right now.
@thsottiaux My rate limits are still not reset.
@Kahsoyas @thsottiaux Same!!!!! But, at 30% until the 23rd.
Don’t think I’ve ever been lower the 60% at any point before.
This needs to be addressed weather it’s real or a mistake, so we can know how to manage. At $200, no one should be guessing and getting caped like this.
@thsottiaux 8 minutes of downtime per month is funny until you realize some developers are now emotionally dependent on a coding agent being available 24/7 to function normally
That shift is bigger than people admit
@thsottiaux It better be a good soul when it gains consciousness
@thsottiaux Please fix mobile bro, Claude code is so much smoother for remote control I just switched over but didn’t realize it was this bad Barely connects to the codex app it just spins on reconnect in ChatGPT mobile app
@thsottiaux 8 minutes per month is a clean way to make 99.98% feel concrete. For Codex, the sharper question is whether those minutes cluster during the exact sessions people rely on.
@thsottiaux Uptime is cute.
The harder reliability metric is: can you hand it a messy task and still get an inspectable diff, stop condition, and rollback path.
Always-on without guardrails just makes mistakes more punctual.
@thsottiaux i can feel exactly which 8 minutes those were