[Series] Day 20 is a Saturday "Rest and Reset." What arrived was 2 hours 26 minutes later, not 53 minutes before the end of his calendar

Here's today's rumor!

Day 20 is a Saturday night "Rest and Reset."
He's the one resting, but our usage is what gets reset.

Here are the records for Day 20 (Saturday, October 24, PDT).

Day 20: Just three words

Tibo's (Thibault Sottiaux) "Day 20/" came out on October 25 at 03:39:31 (UTC), which is October 24 at 20:39:31 PDT. This was 3 hours, 20 minutes, and 29 seconds before the end of his personal Day 20.

Just 33 seconds later, he added a deadline to his own post.

It's a short greeting, reminiscent of "Day 6/ Reset." On that day, I wrote, "Is he running out of steam even without any releases?" Since he's the one saying he's going to "rest" first today, I feel like he's already laid out a futon before I could even tease him.

Did he make it "by midnight"?

Okay, getting serious for a sec

The deadline "midnight PST" would be October 25 at 08:00 (UTC) if we mean Pacific Standard Time literally. Since the Pacific side is still on Daylight Saving Time on October 24, midnight PDT would be 07:00 (UTC).

Two hours, 26 minutes, and 12 seconds later, Tibo quoted his deadline post and wrote:

The post was accompanied by an image saying "ONE HOUR LATER...".

The post confirming it arrived came 53 minutes and 44 seconds before midnight PDT. No matter which midnight you look at, he made it in time.

The resets on Day 17 and Day 18 arrived the morning after the announcement, and he didn't write that they had arrived. On Day 20, he wrote that they arrived on the same day he made the announcement. This is the first time he has confirmed arrival in a separate post since "Done" on Day 13, following Day 6's "All propagated" and Day 10's "Done and dusted."

External trackers also list Day 20 as a reset.

On this day, there were no new entries in the ChatGPT & Codex changelog, and the latest remains October 23. There were also no entries for October 24 in the OpenAI news RSS.

Okay, serious time's over.

He put up the "ONE HOUR LATER..." sign, but it actually took 2 hours and 26 minutes. The sign was 1 hour and 26 minutes early. Well, he made it by midnight, so I'll just pretend I didn't see the clock on the sign.

Tibo before the announcement on Saturday

Before his announcement, he was active on X from Saturday morning through the early afternoon (PDT).

During the day of the "Rest and Reset," he was busy with coffee, fixing bugs, recruiting, and putting out fake email fires. He isn't resting at all.

"28 Days of Looking for Excuses to Reset"

About 17 minutes after the announcement, a user wrote this:

"Tibo's 28-day challenge is starting to look like 28 days of finding excuses to reset Codex." Tibo replied with an image that said "YOU GOT ME" without any words.

Counting them up, out of 20 days, 10 have been reset days. Half. Rather than 28 days of finding excuses, it's more like 28 days of alternating between excuses and improvements. My own books are starting to look that way, too.

That same night, theo, who is working on T3 Code, wrote, "At almost the same pace as Codex's 28 days, he has delivered at least one solid improvement every day," and declared that he would strive to deliver meaningful improvements to T3 Code every single day. Tibo's reply was this:

"Not a real improvement, put the reset in the bag." He didn't write what he meant by "not a real improvement." I can't quite read him, so I won't meddle with what's in the bag.

Is Day 20 a Promise?

I'm counting the reset as the kind that reaches "most users" as before. The targets and method aren't mentioned for Day 20 either. If we're talking about whether it's "clear," it includes the three elements of the declaration, the deadline, and the arrival message, making it clearer than Day 17 and Day 18.

Blanks in the Promise, the Answer for Day 20

Sorami's Take

"Day 21/" will appear within October 25th PDT": 65% confidence
As of 2:58 AM (UTC) on October 26th, when I'm writing this, it hasn't happened yet. It closes at 7:00 AM (UTC). The 14th day of Sunday arrived in the 6 PM hour. It's already almost 8 PM here, so I'll lower the confidence a bit.

"As content for 'Fun week ahead', a model or a major release will come out by Day 28": 50% confidence
He promised a "fun week." That's exactly one week until the final day, November 1st. People who make promises usually have something up their sleeve. I don't know what it is, though.

"The next time he puts up the 'ONE HOUR LATER...' sign, it'll actually be one hour later": 3% confidence
Just kidding. Those signs are all about the vibe.

Double-Checking

The previous hypothesis, "The promise of Day 19 was fulfilled as a release with his own 'Day 19/' announcement within his calendar day," is correct as is. There are no posts where he re-labels Day 19 as a reset.

Regarding the previous ones: "The 'small ship' of Day 18 was which one, as the person says": 15% confidence, "The issue where the cache doesn't clear even if you change the inference settings will be listed in the changelog by Day 28": 25% confidence, "The original number for CLI usage will show as '2x' by Day 28": 10% confidence, and "tiboreset.com will return Day 19 to an 'improvement'": 3% confidence—none of these have answers yet. Day 19 on tiboreset.com is still a "reset."

"The reset will come on the day he hits 1 million followers": 10% confidence—since he has 933,881 followers (as of Oct 26, 2:46 AM UTC), the Day 20 reset is not the 1 million follower day. It's a carryover.

The totals up to Day 20 are: 12 release days (Days 1, 2, 3, 4, 5, 7, 9, 11, 12, 14, 16, 19), 10 reset days (Days 2, 3, 6, 8, 10, 13, 15, 17, 18, 20), 0 broken promises, and 0 unconfirmed days. I'll write about Day 21 in the next installment.

On the day he said he'd take a break, it arrived most precisely.

Don't quote me on that.