Day 5 is "for Pro users"—Predictions for the next word in Codex, dot on mobile, and Day 4's ultrafast was Pro $500

Here's today's rumor!

Day 5 is "next word predictions" for Pro users. Also, you can now create dots from your smartphone.
And as for the scope of ultrafast that I previously called "unconfirmed," the official changelog finally has the answer.

The "Day 5/" post released by Tibo (Thibault Sottiaux) on Day 5 (October 9 UTC) is as follows:

This post quotes the OpenAI Developers official account post from 1 hour and 25 minutes earlier. It states, "As a beta, we've added composer predictions to Codex for Pro users," and "We suggest the next message based on the flow of conversation and how users interact with Codex." This means a feature that suggests your next words in the input field before you even type them.

Two minutes later, he added another "Day 5" post.

The source is the ChatGPT official account post, which says, "You can now create a dot entirely via phone from the iOS and Android ChatGPT apps." Romain Huet also mentions that dots can start working in Codex and follow up on existing threads.

Okay, getting serious for a sec

As of the morning of October 10 (Japan time), the latest entry in the changelog is the October 8 ultrafast update; there are no entries for composer predictions or dots yet.

At the same time as Day 5, 0:12 AM (UTC), Tibo posted "Today, we are announcing ChatGPT." with a link to chatgpt.com. This post doesn't name any new features and doesn't include a "Day" number, so I won't count it.

Okay, serious time's over.

3 Highlights

• For two days straight, improvements came to the higher plans. Day 4's ultrafast was for Pro $500 and some Enterprise/Edu, and Day 5's composer predictions is for Pro. The promised word was that it would "affect most users of codex/work." For those using Plus, it's two days of treats being served through the window.
• To be fair, dots seems to have a wider entrance. It is not stated that the improvement of being able to create from the mobile app is limited to higher plans. Perhaps Tibo's decision to label it as a separate "dots edition" was a plan to have it cover the "most" of this day. That's just my take.
• "No usage consumption" appeared again. Day 2's Auto-review also "doesn't consume plan usage." That's a great way to phrase it for users worried about limits. Though this time, the only ones who can use it are Pro users to begin with.

Filling in the blanks: Day 5 results

Sorami's Take

"Day 5 is a day that won't count toward the promised 'majority of users'": 65% confidence
The main candidate for composer predictions is only for Pro. If dots reaches the majority, that's a different story, but the range by plan hasn't been released yet. Once it is, I'll adjust my confidence level.

"At some point before the 28th, a Pro-exclusive feature will trickle down to Plus": 50% confidence
The only basis for this is that composer predictions mentions "beta." Whether they'll count it as a "new improvement" when it finally trickles down is something I'll laugh about when the time comes.

"Day 6 will be a reset day": 30% confidence
It's been more than two days since the last reset. I'll bump this up slightly from the previous 25%. The basis is simply that some time has passed.

Fact Check

Last time, I wrote regarding ultrafast that "pricing and provision by plan are not written in the posts within my readable range." The answer was in the changelog. Pro $500, and the target Enterprise and Edu.

If that's the case, my previous "Day 4 will be a day that honestly meets the promised 'clear improvement': 70% confidence" is half wrong. Since no plan limitations were mentioned regarding the immediate availability of steering, that part still holds. Ultrafast was "clear," but it wasn't for the "majority." Man, I've been wondering about the price ever since I saw the words "up to 8x." Really.

My previous "Day 5 will be a reset day: 25% confidence" was wrong. My previous "At least one more thing about speed will come by the 17th: 55% confidence" is still in progress.

The cumulative total, with Day 5 finished: 5 days of releases, 2 reset days (the mass reset on Day 2 and the one held for Day 3), and 0 unconfirmed days. The number of release days is the count of "days something was released," not the count of "days it reached the majority." If you count that way, figuring out how many days are left out of 5 is quite an interesting homework assignment.

Day 5 saw an improvement. For Pro users.

Don't quote me on that.