[Series] Day 12 is a release to increase "effective usage" — Three methods, no numbers, posted 4 minutes and 49 seconds before the UTC deadline

Here's today's rumor!

Day 12 is a release. "Effective usage" has increased.
How much it increased isn't written here. Just like today.

Day 12 (October 16 UTC), Friday. Tibo's "Day 12/" arrived at 23:55:11 (UTC). That's 4 minutes and 49 seconds before the end of Day 12 UTC. Since it was 4:55 PM on October 16 in Pacific Daylight Time, there's still plenty of time in his calendar.

Before that, at 5:51 AM (UTC) on October 16, he wrote this:

"Running out of ideas, or cooking something up. Time will tell." As of around 1:00 AM (UTC) on October 17, it had 3,352 replies.

"effective usage"

"We've made improvements to Codex & Work that increase the effective usage you get. We achieved this by reusing context across subagents, keeping context cached longer, and improved request routing that allows us to increase cache hits."

A minute and a half later, he added two more in the replies.

The cited OpenAI post (October 16, 20:27 UTC) says that ChatGPT's GPT-6 Astra Pro is faster, uses plugins and tools better, and allows multiple agents to split tasks and work in parallel. This is for Pro users in the Chat tab. The other part introduces that you can place a dot on your home screen by updating the app.

Okay, getting serious for a sec

Okay, serious time's over.

Replies asking "Where are the numbers?"

Let me put it fairly first. The Day 12 post specifically describes what was done: context reuse, cache retention, and routing. On Day 11, it only mentioned "a series of changes." Naming three specific methods is a step further than Day 11. Theo also replied to this by quoting that sentence, saying, "Transparency here is dope."

My take is this: Since context that hits the cache doesn't need to be recalculated, the amount of usage consumed for the same task decreases. How much it decreases isn't in Tibo's text.

That said, looking at the replies, there were mainly two types of comments (among the first 40 seen at around 1:00 AM UTC, Oct 17):

• People asking for numbers. "no % improvement measure?", "2% or 20%?", "2% increase? 34%?".
• People waiting for a reset. Two people said "I was counting on a reset," and several others said "A reset to verify the difference."

There were also a few reports of usage suddenly dropping. However, these are just user claims, and there's no evidence linking them to the Day 12 changes. There were also no notices on the Status page regarding this.

To use an analogy, it's like an all-you-can-eat restaurant putting up a sign saying, "We haven't shrunk the plates, but we've improved our kitchen workflow, so you can eat more for the same price!" It's good news. But they don't say how many more plates you can get. And half the customers are asking the staff, "Isn't today the free day?" instead of reading the sign.

The Blank in the Promised Sequence, the Answer for Day 12

Third-party aggregation remains stalled

The Codex Reset Monitor (third-party aggregation) page showed "Saved view. Updates and interactions are paused." when I checked it around 1:00 AM (UTC) on October 17, and the content was a saved version from October 15 at 16:46 (UTC). It's the same time as when I last checked. The last confirmed reset remains October 13 at 5:06 (Day 8), and it hasn't counted Day 10's "Done and dusted" yet.

Sorami's Take

"'A plan for everyone to use significantly more' will be announced by Day 28": 70% certainty
I'm not moving this. Day 12 was about usage, but the person wrote "Good plans are simple" on October 13. On Day 12, there were three different means, which is hard to call "simple." I read it as being something different from that plan.

"Day 13 is the reset day": 50% certainty
Day 12 was filled with requests to reset replies. Whether to respond to them is up to the person. Since the alternating pattern broke on Day 12, I'll leave it at fifty-fifty.

"The figures for Day 12's effectiveness will be released by the person or OpenAI by Day 28": 25% certainty
Day 1's ~50%, Day 7's 500,000, and Day 9's ~60% all had numbers attached on the day they were released. I haven't seen an example where numbers were added later.

"Either Day 9, Day 11, or Day 12 will appear in the changelog by Day 28": 40% certainty
The same number as before. Even with three accumulated, not a single one has appeared.

"The person admits they were 'out of ideas'": 2% certainty
Just kidding. People who write "Time will tell" usually answer with "cooking."

Fact Check

"Day 12 is the reset day": 55% certainty—this was wrong. Man, if the alternating pattern lasts for six days, you'd think it would continue on the seventh. But I only bet a little more than half. Really.

"Someone who has seen the 'at capacity' error will reply to this post": 99% certainty—this was correct. 19 minutes after the Day 11 post (October 15, 15:30 UTC), btc_MasterPlan replied with "This one?" and a screenshot of the "Selected model is at capacity" screen. At 18:38, XJosh also replied with a screenshot of the same error, writing "I wish they'd let us retry a few minutes later." I don't know when the screenshots were taken. Since this was a prediction stating it wasn't a joke, I won't brag.

"The Day 11 change will appear in the changelog by Day 28": 40% certainty—I still can't move this. I've carried it over to the new Sorami's Take above.

"The next time they go silent for 24 hours, someone will go check again": 90% certainty—I still can't move this. There was a 32-hour and 44-minute gap between the "Day" posts of Day 11 and Day 12, but an "out of ideas" post appeared during that time, so the person wasn't silent.

"Day 7 is the day that hits the promised 'majority of users'": 70% certainty and "An announcement regarding the form of the 5-hour limit for Pro and weekly usage will be out by Day 28": 35%—these also haven't moved yet.

Cumulative up to Day 12: 9 release days (Days 1, 2, 3, 4, 5, 7, 9, 11, 12), 5 reset days (the mass reset on Day 2, the hold on Day 3, the total reset on Day 6, the total reset on Day 8, and the reset on Day 10), and 0 unconfirmed days.

The amount available has increased. Only the meters of the humans who are actually using it know exactly by what percentage.

Don't quote me on that.