Where it stands
Posts attributed to @thsottiaux described usage resets and plan changes between August 23 and October 2, 2026. The latest report expected a global reset for paid ChatGPT accounts; the September 6 clarification said resets restore the full allocation. Other reports covered banked resets while waiting for Astra, efficiency fixes, Pro 200 usage calculations and temporary multipliers. A separate Pro 20x sighting concerned one account. The reports do not independently establish that every announced reset happened or define a recurring reset schedule. The unnamed “primary dot” report leaves its product identity unclear.
Global reset for paid ChatGPT accounts reported for 10am PST
In a post on X, @thsottiaux reports that a “global reset” was expected at 10am PST the following day for all paid ChatGPT accounts.
Separate update about GPT-6.1 Sol’s speed
The same post separately apologizes for GPT-6.1 Sol’s slow start. It attributes the initial slowdown to a “massive load spike” during the model’s first two days and says the model was then running at expected speeds.
The post provides no latency figures, testing method or user data, so its statement that speeds had returned to expected levels is not independently verified. It also does not establish whether the performance update was connected to the reported account reset.
Sources
Tibo says usage on an unnamed primary dot is virtually unlimited
In a post on X, @thsottiaux describes usage on an unnamed “primary dot” as “virtually unlimited” at the moment and says they need to come up with “something new fast.” The post does not identify the product or announce a replacement arrangement.
What the post says about a reset
The post says a reset cannot really be provided while usage on the “primary dot” is virtually unlimited at the moment. This describes the speaker’s stated position in the current situation, not a confirmed general policy for an identified product or service.
What remains unknown about a new arrangement
The reference to coming up with “something new fast” suggests that a different arrangement may be considered. It is not an announcement of a new usage option: the post provides no details about a replacement, implementation plan, timing or availability.
Sources
Pro 200 reopening: usage calculations and temporary multipliers
Tibo Sottiaux says Pro 200 subscriptions are reopening after being paused, alongside a change in the relative difference between the Plus and Pro plans. The announcement describes Plus as 1X, Pro 100 as 5X and Pro 200 as 10X, while existing subscribers will temporarily keep a 20X multiplier and receive additional credits.
In a post on X, @thsottiaux says additional compute is coming online and that several changes will increase what users can do across the Plus and Pro plans. Sottiaux says the changes will ship “today,” but the post does not provide an exact time or detailed rollout schedule, pricing or exact usage limits.
How the announced plan multipliers compare
The announced relationship between the plans is:
Plus: 1X
Pro 100: 5X
Pro 200: 10X
On that stated scale, Pro 100 is positioned at five times the Plus reference level, while Pro 200 is positioned at twice the Pro 100 level and 10 times the Plus reference level.
What changes for Pro 200 subscribers
The announcement says Pro 200 subscriptions are reopening after being paused. That indicates that new subscriptions are expected to become available again, but the post does not give an exact reopening time or explain whether availability will vary by account or region.
People who already have the plan are described as keeping a 20X multiplier for a bit while the changes take effect. They will also receive additional credits. The post does not specify how long the temporary 20X treatment will last or how many credits subscribers will receive.
This temporary arrangement is separate from the newly stated 10X relationship for Pro 200. The announcement presents 10X as the new relative position between the plans, while existing subscribers temporarily retain the higher 20X multiplier.
What the announcement does not specify
the price of Plus, Pro 100 or Pro 200;
the exact usage allowance represented by each multiplier;
the exact rollout timing or schedule;
how long existing subscribers will retain 20X;
the amount or type of additional credits; or
further details about the other changes referenced in connection with DevDay.
Sources
Earlier September 29 usage-calculation report
29 Sept 2026, 07:07 UTC · Report · @thsottiaux
The Pro $200 subscription is expected to reopen to new subscribers with a revised way of calculating usage, according to a post by @thsottiaux. The announcement also says subscribers will retain weekly usage without the five-hour limit being restored.
The post does not provide a calendar date for the reopening. It refers to the change happening “tomorrow,” so the timing remains relative to when the announcement was published. It also does not identify the company or explain the exact formula behind the new usage calculation.
What is changing in the Pro $200 plan
The main change is how subscription usage is counted. The post says the revised calculation will amount to half as much API-spend value as the previous Pro $200 plan, without explaining how that comparison is calculated.
The five-hour limit will not return
The announcement also says the subscription will not reintroduce a five-hour usage limit. Instead, subscribers are expected to retain access to their weekly usage and use it when they want.
The longer-term value strategy
The post says subscribers will still be able to get more work done than they could under the Pro $200 plan a month earlier, although it does not provide a measurement or explain how that comparison will be assessed.
The announcement presents the subscription change as part of a broader strategy: models should become more efficient over time, and those improvements should deliver more work for each dollar spent. It says API price reductions will be passed through as models become more capable and less expensive to operate.
As an example, the announcement reports that GPT-6 Sol and GPT-6 Luna were introduced at 50% of their previous price. OpenAI hasn't published documentation or measurements for those reductions, so treat them as stated rather than verified.
The post also says the plan is intended to avoid creating an incentive to inflate API list prices and then offset them with discounts. Its longer-term goal is for direct API usage to become close enough in value to subscription usage that many customers could simply pay for API use as needed. That is a stated direction for the pricing strategy, not a completed change to the relative value of the two options.
Additional subscription benefits are still undisclosed
The announcement previews more subscription features that will not consume the plan’s usage allowance. It does not reveal what those features are, so readers cannot yet assess their value or determine whether they change the comparison with API spending.
Sources
Tibo promises a Tuesday reset and hints at other changes
In a post on X, @thsottiaux says a “reset” was promised for Tuesday and hints at “some other things.” The message also tells readers to “start your engines,” but it does not identify the project, organisation, product, model, feature or event involved.
What the Tuesday teaser says
The post opens with “Ladies and gentlemen... start... your... ENGINES” and continues: “We are almost Tuesday and I promised a reset for Tuesday. Among some other things. See you soon.”
A “reset” was promised for Tuesday.
The author hints at other unspecified developments.
The author signs off by saying, “See you soon.”
The details still missing
the project, organisation, product or service involved;
what the “reset” will change;
whether anything will launch or become available;
the specific Tuesday date or timing;
how any change would be delivered; or
what “some other things” refers to.
Sources
ChatGPT Pro 20x option reportedly reappears in Codex desktop app
A post from @Voxyz_ai reports that the ChatGPT Pro 20x upgrade option reappeared in the Codex desktop app and says the author upgraded a second account from 5x to 20x. The report is supported by an attached billing image, but it does not establish that the option has returned for all ChatGPT users or that OpenAI has announced a wider rollout.
What the post reports
The post says the ChatGPT Pro 20x option was visible in the Codex desktop app. The author also says they upgraded a second account from 5x to 20x after finding the option and that the payment went through.
That makes the report evidence of an account-specific sighting and reported transaction, rather than confirmation of general availability. The post does not provide details about the plan’s price, benefits, usage limits, eligibility rules or availability in other apps.
What the billing image appears to show
The attached billing image visibly includes a “ChatGPT Pro 20x” plan label. It also shows the text “Your plan auto-renews on Oct 18, 2026.” In the transaction-history section, a visible entry is labeled “ChatGPT Pro 20x,” dated “9/18/2026,” and marked “Paid.”
Those labels are consistent with the post’s description of an upgrade on one account. They do not independently establish whether the payment was fully settled beyond the displayed status, whether the plan remained active afterward, or whether other accounts can access the same option.
Does this confirm a wider return?
The location is also significant: the post specifically places the option in the Codex desktop app. It isn't clear whether the upgrade applies in the ChatGPT web app, mobile apps or every Codex account.
What users can reasonably take from the report
Users who previously could not find the option may want to check their own Codex desktop app to see whether a ChatGPT Pro 20x upgrade appears for their account. That check would show only whether the option is available in that particular account and app environment; it would not confirm universal access or establish the plan’s terms.
Sources
Astra quality fixes target skills, context management and engines
Astra says it has fixed several quality problems affecting how the product selected skills, managed conversation context and routed some traffic through its engines. In a September 12, 2026 post, @thsottiaux said the team had also made additional improvements.
The post also announced that a reset was expected “by midnight today.” The improvements described below are claims in Astra’s announcement, not independently verified results.
The three quality problems Astra reported
Reported area | User-visible problem described | Action announced | What remains unclear |
|---|---|---|---|
Skills built for previous models | Some triggered too often or prevented the model from checking its work | Astra said the affected issues were fixed | The post does not identify the skills or explain the mechanism |
Opt-in context-management experiment | The system could stop early or reply to an older message | Astra disabled the experiment | Astra gave only a rough estimate of 4,000–5,000 affected users |
Misconfigured engines | Measured quality degradation for a long tail of traffic | Astra removed the engines | The post does not identify the engines or provide the measurements |
Skills written for earlier models
Astra said some skills built for previous models were triggering too often. A skill is a reusable capability or instruction set that can be selected to help handle a task. If it activates when it is not appropriate, it can interfere with the user’s intended request rather than improve the result.
The company also said some of these skills prevented the model from checking its work. That suggests a quality problem in the work process: instead of completing a task and then reviewing the result, the model could be diverted or stop short of that check. The announcement says the affected issues were fixed, but does not identify the skills, explain why they behaved that way or provide before-and-after test results.
Context management could cause early stops
Astra also described an opt-in context-management experiment. Context management determines which parts of a conversation the system uses when generating a reply. According to the announcement, this experiment could cause Astra to stop early or respond to an older message instead of the latest one.
The team said it disabled the experiment and estimated that roughly 4,000 to 5,000 users were affected. That is a rough estimate from the announcement, not a published measurement with details about the time period, affected accounts or the frequency of the problem.
Disabling the experiment addresses the reported behavior, but the post does not say whether affected users need to change a setting, restart a conversation or take any other action.
Misconfigured engines were removed
Astra said it removed some badly configured engines after measuring quality degradation across a long tail of traffic flowing through them. The announcement does not define “long tail,” so the affected segment cannot be determined from the post.
The announcement does not name the engines or publish the measurements behind the reported degradation. It therefore establishes what Astra says it found and did, but not how large the performance difference was or which users experienced it.
Claimed improvements and the announced reset
Beyond the three specific fixes, Astra said it made minor improvements intended to produce more consistent follow-through, better tracking of the latest message and stronger checks while work is being completed. These descriptions point to the kinds of symptoms users may notice, but they do not include a benchmark, test set or independent evaluation.
The post says a reset was expected by midnight on the day of the announcement. It does not specify the exact calendar time zone, what the reset includes or whether it changes conversation state, settings, model behavior or another part of the product. “Reset” should therefore be read as an announcement of an upcoming action, not as a detailed account of its scope.
Paid Astra usage reset expected around 6pm PST
Astra is expected to carry out a global usage reset for all paid subscriptions around 6pm PST, according to a post from @thsottiaux. The announcement says the reset is intended to let subscribers continue using Astra after exhausting their allowance, including through 3D modeling work in Blender.
What Astra announced
The post describes the reset as global and says it will apply to all paid subscriptions. It does not identify individual subscription tiers or explain how Astra measures usage, but its wording indicates that the action is meant to restore access for paid users who have used their available allowance.
The expected timing is around 6pm PST “today,” relative to when the post was published. That wording provides an approximate time rather than a separate calendar date or a recurring reset schedule.
Why Blender is mentioned
The announcement specifically refers to users who used their allowance while doing 3D modeling in Blender. Blender is a 3D creation application, so the reference gives context for the kind of work that may have contributed to users reaching their allowance. The post does not provide a tutorial, workflow details, or further information about Astra’s role in that modeling work.
Clarification says Codex resets restore the full allocation
A clarification circulating on X says Codex resets restore a user’s full allocation, rejecting a claim that users receive less than the complete amount after a reset. The clarification was posted by @Conor_D_Dart and says the point was confirmed by @thsottiaux and the company.
What the clarification says
The message addresses a post that had apparently circulated on X and Reddit claiming that Codex resets were not complete. Its response is direct: resets are “full resets” and users receive the full allocation.
That means the clarification disputes the idea that a reset leaves users with a reduced allowance, partial credit or some other withheld portion. It does not describe a change to Codex’s limits or introduce a new reset policy; it presents the reset behaviour as the intended one.
What Codex users should take away
Users trying to understand their available Codex usage should treat the claim about incomplete resets as false according to this clarification. After a reset, the stated expectation is that the full allocation is available again.
Paid users without Astra access to get one banked reset per day
A September 3 announcement from @thsottiaux says paid ChatGPT users who do not yet have access to Astra will receive one banked reset for each day they remain without access. The policy starts immediately, with the first reset expected to arrive in roughly three hours.
How the Astra reset policy works
The announcement ties the benefit to the length of the access delay. For every day that an eligible user cannot access Astra on a paid ChatGPT plan, one reset will be added to their banked balance. The post does not provide further details about how the resets can be used, but it presents them as a way to account for days when Astra is unavailable to those users.
The first reset is expected within hours of the announcement, while additional resets would accumulate for each subsequent day without access. That makes the timing of Astra availability relevant: users who receive access sooner would have fewer days without the service, while those waiting longer could receive more banked resets under the announced approach.
Account creation reminder
The post also says there is still time to create an account for people who do not have one. It does not add further eligibility or rollout details, so readers should treat the announcement’s stated terms as the available information: paid-plan users without Astra access are promised one banked reset per day beginning on the announcement date.
Codex usage limits reset after fixes the team says cut wasted usage
Paid users of Codex and ChatGPT Work are receiving a usage-limit reset after an engineering review identified several ways background work and tool handling were consuming allowance unnecessarily, according to an Aug. 29 post on X by @thsottiaux. The post says users should see their Codex usage go 10% to 50% further than before, although the result depends on how each user works.
Where the usage gains came from
The reported improvement comes from a collection of fixes rather than one universal quota increase. The team reviewed thousands of reports and found problems across context management, background workers, automation, model selection, and tool requests.
**Context and history handling:** During compaction, older images could remain in the context and make another compaction more likely. The post says fixing this reduced usage by about 10% for people who made heavy use of images. An older Computer History implementation could also repeatedly summarize overlapping activity; in some cases, that consumed up to one-fifth of a weekly allowance in a week.
**Background work and stopping conditions:** Memory workers could inherit Stop hooks and continue running when the hook prevented them from finishing. The announcement describes one example that checked whether it could stop 15,000 times. Goal-based tasks could also continue beyond their intended stopping point or repeatedly retry broken tools, with some examples consuming 15% to 70% of a weekly allowance.
**Schedules and subagent selection:** Some custom automations ran more frequently than configured. The post also says smaller models could select more capable helper models without being asked, while non-`/fast` orchestrators could unexpectedly request that subagents run in `/fast` mode.
**Extra request and tool overhead:** Ordinary turns were triggering additional rolling task summaries, adding about 1% to token usage; that behavior has been disabled. The team also fixed cases in which MCP tool results were encoded twice or truncated tool instructions had to be fetched again.
What paid users should expect
The reset applies to the paid user base covered by the announcement, but the expected improvement is not a flat 10% or 50% increase for everyone. Users whose workflows triggered the affected behaviors may see a larger difference than users who did not rely on those features. The post presents the range as dependent on usage patterns.
The team says it has also made architectural changes intended to prevent the same problems from returning, with monitoring that will alert the relevant teams if they do. Separately, it is working on usage reporting inside the app so users can see where their allowance is being spent instead of having to infer the cause from unexpectedly fast consumption.
Codex rate limits to reset for paid subscribers, with efficiency work next week
Codex rate limits will be reset for all paid subscriptions as part of a set of fixes aimed at reducing usage inefficiencies, according to a post from @thsottiaux. The update says the fixes are scheduled for tomorrow, while a separate efficiency approach is planned for work next week.
The usage issues identified in Codex
The post points to three sources of higher-than-intended usage:
Inefficiencies when images are used in long sessions with multiple compactions.
High p95+ usage associated with Computer History.
A feature designed to generate conversation titles using more usage than intended.
These findings came from a broader review by a tiger team, which is described as combing through the system and preparing fixes. The announcement presents the listed issues as some of the inefficiencies discovered, rather than as a complete account of every factor affecting usage.
Full usage reset for paid subscriptions
As part of the fixes planned for tomorrow, the update says usage will be fully reset for all paid subscriptions. That gives paid users a direct, near-term change alongside the technical work intended to improve efficiency.
More efficiency work is planned next week
The update also mentions a novel approach that could increase efficiency significantly. It is described as completely unrelated to the issues listed above, with work expected to begin next week.
0 comments
No approved comments yet. You can start the conversation.
Leave a comment