[Series] Day 17 is "Reset Day" 4 minutes and 38 seconds before closing. It says "By 10 AM tomorrow," but everyone disagrees on when that "tomorrow" is

Here's today's rumor!

Day 17 is "Reset Day," arriving 4 minutes and 38 seconds before closing.
It says "By 10 AM tomorrow." But when exactly is that "tomorrow"?

This is the record for Day 18 (Thursday, October 22 UTC). I'll start with the answer to Day 17, which I had put on hold last time.

Day 17, just before the date changed

Tibo's (Thibault Sottiaux) "Day 17/" came out at 06:55:22 (UTC) on October 22. In Pacific Daylight Time, that was 23:55:22 on October 21. That is exactly 4 minutes and 38 seconds before the 17th day of his own calendar closed.

"Bet you thought we forgot, but no, it’s reset day today." 32 seconds later, he replied to his own post.

If you ask me if I thought he forgot, I'd say no. Last time, I bet 75% certainty that it would come out within October 21 PDT. I was right, with 4 minutes and 38 seconds to spare. Seriously, that was a heart-stopping way to get it right.

Counting by his own calendar, the announcement times are as follows:

After a series of 6 PM announcements, Day 17 was a sudden rush right before midnight.

"Daybreak Blue" is the entry point for OpenAI's cyber defense initiative, Daybreak. On X on August 10, OpenAI explained that it is a gateway for many defenders, offering access to state-of-the-art models, including GPT-5.6 Sol, with safety measures for defense. As for the "it only takes 15 minutes" part, that's according to Tibo.

An announcement for a disaster drill following a reset notice. It's like being in a shop right before closing and being told, "Everything is free tomorrow. Also, please check the evacuation routes."

When is "10 AM tomorrow"?

Okay, getting serious for a sec.

The reply was posted at 23:55:54 on October 21 PDT. From that moment, "tomorrow" is October 22. 10 AM in PDT is 17:00 (UTC).

He wrote "PST" (Pacific Standard Time). The Pacific coast is still on Daylight Saving Time on October 22; it returns to Standard Time in the early hours of November 1. If he literally meant 10 AM PST, that would be 18:00 (UTC).

Third-party trackers are reading this reply differently. All three are third-party counts.

There are no posts from Tibo stating that anything arrived during October 22. His only posts on this day are a short reply at 07:00:03 and a reply at 22:58:07.

Therefore, we haven't been able to primary-source confirm whether the Day 17 reset actually arrived. If the 17:10 timing from codex-reset.com is correct, it arrived 10 minutes past 10 AM PDT, or 50 minutes before 10 AM if we follow literal PST.

Okay, serious time's over.

Thanks to the "PST" label, a 10-minute delay turns into 50 minutes of breathing room. Time zones can be a convenient excuse depending on how you use them. I'll keep that in mind too.

By the way, November 1 is Day 28, the promised final day. The only day his "PST" will truly be accurate is on that final day.

How to count Day 17?

He announced "Day 17/" and said the content is a "reset." Since this series has always counted days according to his own announcements, I will count Day 17 as a reset day.

However, for resets from Day 6 onwards, he had always stated they would arrive within the same PDT day as the announcement. Day 6 and Day 8 were "EOD," Day 10 was "within 2 hours," Day 13 was "within today," and Day 15's announcement itself was "and reset." This is the first time in this series where the announcement day and the arrival day were split. I think tiboreset.com split them because that's why they counted the arrival as Day 18's reset.

Count by the announcement or count by the arrival? The day shifts depending on how you keep the books. I learned the hard way last time about how books are closed, so today, I'm going with his announcement.

Day 18 is still ongoing

Here is the timeline for Day 18 (October 22, UTC).

The Decisions API is for developers using the API. The Uber news is about a single company's adoption, and the METR news is about safety evaluation. Neither can be called an improvement for "the majority of users of codex/work." Also, the answer to the riddle from last night, "Dot, Dotter, Dottest," still hasn't appeared.

At 01:00 UTC on October 23, when I made this table, "Day 18/" does not exist. October 22, PDT, doesn't end until 07:00 UTC on October 23, so there's still over 5 hours left. Day 17 arrived with 4 minutes and 38 seconds to spare before it closed. I won't go as far as to say "it didn't come" here, no need to repeat last time's correction.

The blanks in the promises: Answers for Day 17 and Day 18

Sorami's Take

" 'Day 18/' will be released within October 22, PDT": 70% confidence
The announcements for Day 14, 16, and 17 all came after sunset PDT. The reason I haven't seen it yet is simply because it's not evening there yet. Still, those 4 minutes and 38 seconds from Day 17 really gave me a fright.

"Day 18 is a Release Day": 60% confidence
After the Day 13 reset came the Day 14 release; after the Day 15 reset came the Day 16 release. We've seen this pattern of "release the day after the reset" twice now. I'm reading this as the third time it will continue.

"Tibo will write that the Day 17 reset has arrived": 55% confidence
This is someone who often drops a quick word after things arrive, like "All propagated" on Day 6, "Done and dusted" on Day 10, or "Done" on Day 13. However, since some counts already show it has arrived, they might just move on to the next thing without writing anything.

"The announcement for the final day on November 1 will again say 'PST', and only that day will be correct": 3% confidence
Just kidding. But if they do, I'll be the only one clapping that day.

Checking the Answers

The previous " 'Day 17/' will be released within October 21, PDT": 75% confidence—is a hit. 4 minutes and 38 seconds left. I'm feeling pretty smug about that. Though, my lifespan might have shortened just a little bit because of it.

The previous hypothesis "The promise for Day 17 will be fulfilled within the individual's October 21 via their 'Day 17/' announcement" is also a hit.

The previous "The content of 'Day 17/' refers to the nighttime 'Dot, Dotter, Dottest'": 60% confidence—is a miss. The content was a reset.

The previous "The content will be an upgraded version of 'dots' or something that increases usage": 55% confidence—is also a miss. Well, a reset is kind of like increasing usage, so I'd like to say the second half was correct. I'm just saying. Since the premise of "dots" was wrong, I'll count it as a miss.

The previous "Another 'Day N/' will come within 10 minutes of me closing the books": 3% confidence—is a miss. This time, it was more than 5 and a half hours away. I'm glad I missed a joke-based prediction.

The previous-previous "Day 17 (October 21, UTC) is a Release Day": 55% confidence—is a miss. It was a Reset Day.

"The second installment of COOKED will be out by Day 28": 45% confidence; "A reset will come on the day they hit 1 million followers": 10% confidence; "Either Day 9, 11, or 12 will appear in the changelog by Day 28": 30% confidence—these all remain unchanged.

The cumulative tally, up to partway through Day 18: 11 Release days (Days 1, 2, 3, 4, 5, 7, 9, 11, 12, 14, 16), 8 Reset days (Days 2, 3, 6, 8, 10, 13, 15, 17), 0 unfulfilled promises, and 1 unconfirmed day (Day 18; not certain until 07:00 UTC on October 23, when October 22, PDT, ends). I'll count the Day 17 reset based on the announcement and verify if it arrived next time.

The promise was kept with 4 minutes and 38 seconds to spare. As for the arrival time, whether it's considered late or early depends on which clock you're watching.

Don't quote me on that.