Updating 'Chat' while leaving Codex and Work unchanged: Day 3 brings GPT-6 and the distribution of banked resets.

Here's today's rumor!

The big one for Day 3 is GPT-6. It's heading to Chat.
Codex and Work remain unchanged. That's what the official word says.

Tibo (Thibault Sottiaux) posted the "Day 3/" update on October 7th (UTC), and here it is:

The main event is GPT-6 in Chat, but since it's also a day to celebrate reaching a record 40 million active users across Codex and ChatGPT Work, the gist is that everyone with a paid account is getting a "banked reset" (summary). According to this reply, the distribution will be completed by the end of today Pacific Time.

Okay, getting serious for a sec

OpenAI has announced the GPT-6 details via their official account and announcement article. The "Scope of Provision" section in the announcement article states the following:

A "banked reset" isn't a new term. In the June 11th entry of OpenAI's ChatGPT & Codex changelog, they mentioned adding a feature for Plus and Pro users to bank their usage limit resets, providing one free reset at the start. It essentially means a reset that users can choose when to use.

As of 10:00 AM October 8th (JST), no post has been found confirming that the distribution of banked resets is complete. The third-party Codex Reset Monitor also hasn't counted this banked reset, still showing the latest reset as the Day 2 batch from October 7th at 3:35 AM (UTC).

Okay, serious time's over.

Four Key Highlights

• The main event falls outside the promise. The promise was "clear improvements affecting the majority of Codex/Work users." Regarding the Day 3 highlight, GPT-6, the announcement itself says, "Work and Codex models will remain unchanged." They're being honest. Honestly, maybe too honest—it doesn't quite fit the "promise" category, does it?
• Instead, we got resets. The promised wording was "one improvement or a full reset." Day 3 is a combination of a big release outside the original scope and the banked resets. The reason for the resets wasn't because they ran out of stock, but to "celebrate 40 million users." Even if the reason differs, it's still a reset for those receiving it.
• Celebratory resets have precedents. According to Codex Reset Monitor records, resets were distributed in July at milestones of 6 million, 8 million, and 9 million users (third-party records). From 9 million to 40 million in just under three months. If they give them out every time there's a milestone, every day's going to be a milestone soon.
• There are two different numbers. Looking at the Japanese version of the announcement page, the speed for starting answers for questions requiring web search was described as "32% faster on average" in the diagram explanation, but "44% sooner" in the main text. Two different numbers for the same comparison on the same page. Within what I could see, there's no explanation as to which one is correct.

Day 3 Answers to the Day 1 Blanks

Sorami's Take

"The day 3 promise was fulfilled by the reset, not the shipment": Confidence 75%
The main event, GPT-6, is explicitly excluded from Codex and Work by the announcement itself. What falls under the "majority of codex/work" promise is the banked reset. The weak point is that the person in charge wrote that the Day 3 highlight is "GPT-6" and called the resets a "celebration." In their calculations, this is probably a shipment day.

"The banked reset counts as a 'full reset' as per the promise": Confidence 55%
Since usage limits are supposed to return when used, the content is indeed a reset. However, unlike Day 2 where everyone's limit reset all at once, everyone chooses their own timing. I think opinions will split depending on how you interpret the phrase "full."

"On Day 4, the banked resets from Day 1 will also arrive": Confidence 2%
Just kidding. But milestones don't wait for anyone.

Fact Check

My previous prediction, "Day 3 will be carving out 3.1 to 3.8: 5% certainty," was wrong. Day 3 was just a single Day 3/—no decimal points anywhere. Man, I really thought the decimals on Day 2 were a one-off performance from the start. Seriously.

My previous prediction, "His 'Day 1' lasts until tonight PST: 60% certainty," was close to being right. They've written the reset deadline as "EOD PST." Since the official declaration of the cutoff hasn't happened yet, I won't call it a confirmed hit. But my reading was on the right track, wasn't it?

So far, after Day 3 is over, we have 3 days of shipments, 1 day of resets, and 0 unconfirmed days. For the Day 3 reserved resets, I'll count them in the fact check once it's confirmed they've finished distributing them. As for the Day 3 shipments, I've verified them via OpenAI's official account and news articles.

The main event went outside the promise. The promise, however, was kept by the reset.

Don't quote me on that.