🏅 Top traders

Note that the comment on 27 March is about 26 March.

The per day page shows NO outage for 27 March.

https://status.openai.com/uptime/tbwycthtqm69

At the moment, based on the description, I am still more inclined to count this as a YES. It is a lasting record of the resolution of an incident, timestamped 3/27.

Updated to show partial outage on 26th but shown as resolved under 27th at present. Wonder if that will move back to 26th?

@ChristopherRandles I don't think so, but as they say in the message, they have updated the status chart to show an outage yesterday, which means I messed up on this one, and in general may need to wait a full day before resolving

@WalterMartin if the 26th was incorrectly resolved, do you want me to unresolve it?

@chrisjbillington yeah, given the change on their end it should resolve to yes. Thank you!

@WalterMartin it's unresolved, you can re-resolve

ChatGPT was down for about five minutes just now, for me and at least some others (according to a Twitter search). It's working again, not sure if they'll report an incident for it.

@chrisjbillington I guess they don't care. Selling!

"Elevated errors in ChatGPT" incident on the 25th now marked as resolved. 25th resolves YES @WalterMartin?

@chrisjbillington We'll give it a little more time to be sure. I know it seems silly, but they have made changes after the fact before. I'll certainly close it out in the next 12 hours.

Well there isn't an incident reported but ChatGPT is down for me at the moment, so expecting there might be one

@chrisjbillington different now. Also down for my girlfriend, so not just me. Though the "MEL" after the hash above perhaps suggests it's a problem with local servers (I'm in Melbourne)

April market is now up:
/WalterMartin/will-openai-report-an-incident-toda-6edae05210f4

Time to resolve? An incedent has been reported

@CyfralCoot It might end up saying this incident was resolved on 21st so not clear which way 20th resolves yet.

@ChristopherRandles its already reported "Unresolved incident: Elevated errors in requests to GPT-4 in the API.". What they gonna do, remove it?

@CyfralCoot Have you read the description? If the resolution is on 21st, "on which day the lasting record is left"?

The related may stay on both or maybe not?

@ChristopherRandles thanks, you're right on - @CyfralCoot, we've seen a little bit of weirdness in the past where they have ended up putting an incident that started, e.g., on 3/20, fully on 3/21. Given the stated goal of having resolutions of this market line up with the "historical record", I'm going to wait to resolve 3/20 until the dust has settled.

Believe me, I want my profit from this market as well, but I can't in good conscience resolve yet.

@WalterMartin to be clear, it is still very likely that this will resolve to true - even if the incident gets put on the 21st, the daily color chart will still likely show something from yesterday.

The daily color chart does in fact show a record from yesterday even after incident closeout.

@traders, I don't want there to be any surprises here - based on the criteria I set out originally, it's looking decently likely that there will be a lasting record of the ongoing incident today, at least in the daily record for API and ChatGPT. We current see "degraded performance" there, and if you hover over the green bars for 3/15, there is a "related" section referring to the ongoing incident. Please plan accordingly.

Resolves as NO, they've opened up the "March 12, 2024" section now.

Resolved yes for incident that stretched 12 seconds into 8th before being stated as resolved but normalised over 4 hours earlier. ho hum.
I would have thought that may end up on Mar 7 only.

@ChristopherRandles Based on the specific resolution criteria I laid out (and Calvin clarified), we have a lasting record of the incident on 3/8. I'm not a big fan of it either, but this is how I promised to resolve the market ahead of time.

FWIW, it's 12 minutes tho

@WalterMartin Ah yes 12 minutes.

We had one incident where it said resolved ~8am the following day but later this was adjusted to ~23:56 of the same day as most of the incident. This resolved as the earlier day. I had skim/read the description message as meaning if such a situation occurs again you wait rather longer to see if the resolved time is pushed back so the lasting record is put on the earlier day.

Now I read it more carefully, I see it isn't saying that. You are using the first resolved time posted which isn't what happened on that earlier incident before that description clarification was added. So apologies.

Perhaps worth posting this as the description still doesn't clearly say what happens if one resolved time is put up and it is later changed.

@ChristopherRandles Good point. Probably I should wait a bit longer before resolving to reduce the chance of running into that scenario.