Codex Day6 usage reset: Tibo says by EOD—check Usage first

Day6 confirms a reset. Why does that not confirm receipt?

At 22:15:39 UTC on October 10, 2026—06:15:39 Beijing time on October 11—Tibo posted “Day 6/ Reset.” Forty-one seconds later, he added “Global reset by EOD” in the same thread, saying that the global reset would happen by EOD. Day6 is the author’s own numbering.

EOD stands for end of day and usually means before the day ends. Neither of these posts specifies a timezone or exact time. The date displayed on your phone does not establish the deadline for a global reset.

What is confirmed is the reset announcement and the wording about completion timing. Plan coverage, the particular usage windows, whether it is applied automatically or stored as a benefit, and when each account receives it have not been fully explained. The word global does not replace a published eligibility list.

Check Usage before deciding whether to resume a task

First, confirm the account, workspace, and product you are signed into. In Settings → Usage, check remaining usage, replenishment times, and available reset counts. Save the observation time and timezone to establish the current state; do not launch a large task just to test whether the reset arrived.

Second, if you have a record from before the reset, compare the same account and usage window. Without an earlier record, you can only confirm what is available now. An official announcement, someone else’s screenshot, and your own Usage page are three different kinds of evidence; none substitutes for another.

Third, once you have enough usage available, check the original task’s run records and actual output. If its outcome is unknown, resume checking that same task first. Replenished usage does not mean a failed task was rerun for you, and does not justify resubmitting a task that could perform external actions.

A hypothetical example: you resume work after seeing the reset announcement, then hit a limit again. That only shows that insufficient usage remained at that point. Comparing the times of the announcement, your Usage observation, task start, and later limit can help distinguish a reset not yet applied from usage consumed by new work. This site has not tested the example with a real account.

Is Day6 the same as the Day3 banked reset?

The Day6 posts have not explained whether the reset is applied automatically or stored, so do not assume it works like Day3. Tibo confirmed on October 8 that the earlier Day3 banked reset for paid accounts had arrived. Banked means the reset benefit is stored for use when needed; it does not mean current usage is automatically cleared.

Check your available reset count and expiration information in Settings → Usage. The original 28-day plan posted on October 4 promised a meaningful improvement or a full reset each day. The two Day5 capability updates did not announce a new reset at that time; that does not imply there would be no later resets.

The general mechanism can refresh eligible five-hour or weekly windows, consuming one reset only when at least one window is successfully refreshed. Pro100, Pro200, and Pro500 currently have no five-hour limit. A common expiration date for this particular Day3 grant has not been verified; follow the benefit shown in your own Usage page.

Day5 predictions are free. What about the task you send?

Day5 introduced composer predictions: suggestions for the next message you might want to send, based on the current conversation and your writing style. OpenAI Developers described a Pro beta for local tasks in the Codex desktop app. That does not establish availability for Plus, every plan, or every client.

Free applies to generating the suggested input itself. It does not make the Codex task you send afterward free, and did not announce a reset of previously consumed usage. Whether to execute a task and how it consumes usage still depend on the particular task, mode, and account benefits.

First, when an input prediction appears in an eligible local task in the Codex desktop app, you can press Tab to accept it. The prediction is candidate text suggested from context, rather than a new requirement you have already approved.

Second, after accepting it, check its scope and wording, then edit and send it. Accepting candidate text and sending it are separate actions. Free predictions do not waive usage for subsequent tasks, and the Day6 announcement does not turn them into unlimited work.

Third, if you do not want predictions in the input box, you can disable the capability in Settings. The official reply confirms this control, but the posts read do not provide a complete version list or rollout times for each region.

Creating and messaging Dots on mobile still depends on plan and region

ChatGPT officially confirmed that Dots can be created on iOS and Android phones. Current eligibility documentation lists Pro100, Pro200, and Pro500 for users over 18 outside the European Economic Area, the United Kingdom, and Switzerland. This does not mean every Pro region has access, or establish Plus eligibility.

Business Premium is rolling out gradually worldwide. Enterprise is also rolling out gradually worldwide, but is disabled by default and requires a workspace administrator to enable it. Even an eligible plan may not show the entry point yet. Mobile functionality, plan eligibility, and access on your actual account need separate checks.

Conversations with a dot do not count toward ChatGPT usage limits, but the Work and Codex tasks it starts or manages still count toward the relevant product’s usage. The documentation also mentions a plan allowance for deeper work and expanded limits during the first launch month, without giving a universal number on the pages read. That is not unlimited free work.

First, use the latest ChatGPT mobile app, open dots, and complete the introduction. The official post confirms iOS and Android. Current documentation does not support mobile web and does not specify a universal minimum app version.

Second, name your dot and choose its appearance. Connecting plugins for email, calendars, files, and similar services is optional: you can skip them and add them later. Creating a dot does not authorize every app.

Third, message the same dot in the app. Tibo’s phrase text dot refers here to text conversation, rather than proof that phone SMS has launched: the current official Channels documentation still says Texting coming soon. You can continue using the same dot in the mobile app, desktop app, and desktop browser.

ChatGPT’s official reply also says a dot can start Codex work, continue an existing thread, and draw on ChatGPT conversations, Codex threads, and automations. It can also inspect or edit Scheduled Tasks in ChatGPT Work. Delegation does not imply access to every historical conversation or permission for arbitrary tasks.

To let a dot work on your own computer, you can connect it separately. Local work requires the computer to be online and the ChatGPT app to be open. That connection is optional, and the dot also has a cloud computer. Messaging channels, app plugins, and computer access are separate permissions; sending a mobile message does not automatically grant access to your local computer or email.

Day4 steering and rendering improvements do not speed up every part of a task

At 19:15:14 UTC on October 8, Tibo called instant steering and GPT-6.1 Sol Ultrafast Day4. Steering means adding requirements while a task is running to adjust what happens next. An immediate response does not mean completed external actions can be undone, or automatically switch the task to a more expensive mode.

The Day4 post did not list the full plan, client, and regional scope for steering. Do not apply Ultrafast’s Pro500 eligibility directly to steering. After adding instructions, check whether subsequent explanations and actions actually return to the revised scope.

A separate reply at 05:25:16 UTC on October 9 addressed slow feedback within the Playground’s client-side and network scope. Tibo said optimized rendering had shipped. Rendering draws received output on the page; it is distinct from model generation and tool execution, and does not establish that every slow Codex task has been fixed.

Ultrafast accelerates token generation. The official launch post claims up to eight times the speed of Sol Standard. That ceiling is not a fixed speedup for every task; networking, tool calls, waits for authorization, and later corrections still affect the total time taken.

How do Ultrafast eligibility and the 8× and 6× charges differ?

Verified subscription eligibility at launch was Pro $500 and eligible Enterprise and Edu plans, for Codex and ChatGPT Work. Enterprise is disabled by default; workspace owners must enable it for selected users or the whole workspace. Enterprise agreements using credits or dollar-based usage billing, and Edu plans using credits, qualify. Legacy Enterprise plans based on rate limits do not.

Other self-serve plans did not receive Ultrafast access at launch, even if they purchased credits. Composer predictions being available to Pro does not give the same account Ultrafast access. Eligibility for one capability does not substitute for eligibility for the other.

Ultrafast consumes included plan usage at eight times the Standard rate. Purchased credits and Enterprise usage billing are charged at six times the rate. Pro500 uses included usage first, then available credits after that runs out; Work and Codex share usage. This is a different billing stage from free input predictions.

Check usage, eligibility, and data residency for the speed mode separately. Faster does not mean unlimited concurrency, and buying credits does not change plan permissions. Do not transfer Dots’ regional conditions or the free status of input predictions to Ultrafast.