On October 5, 2026, Tibo Sottiaux, who runs Codex and ChatGPT work at OpenAI, posted a pledge with a deadline attached: over the next 28 days, "each day we'll either ship one thing that is a clear improvement and relevant for most codex/work users or ship a full reset."
That is an unusual promise for a developer tool. It converts a vague "we hear you" into a daily scoreboard, and it puts the most-complained-about part of Codex, usage limits, inside the product roadmap. This post covers what he actually committed to, why it arrived now, how to plan your own usage around it, and what it does and does not change about your Claude Code vs Codex decision.
TL;DR — the questions people are asking
| Question | Short answer |
|---|---|
| What was promised? | Every day for 28 days: one clear improvement for most Codex/work users, or a full usage reset. |
| Is that 28 resets? | No. It is either/or. Some viral summaries said "28 resets"; Tibo did not write that. |
| Why now? | Weeks of quota-cut complaints, a halved Pro 200 plan, a rough GPT-6.1 Sol launch week, and Claude Opus 5.5 winning developer mindshare. |
| What is the focus? | "Simplifications, more efficiency for more usage, groundbreaking features or new models." |
| Does it cancel the Oct 30 Pro 200 cut? | No evidence of that. The 20x to 10x change is still scheduled. |
| Should I hoard quota for a reset day? | No. Resets typically replace the leftover bucket. Spend it. |
| Should I switch to Claude Code? | Do not switch wholesale on one thread. Keep both harnesses working and measure. |
What Tibo actually said
There were two posts, roughly a day apart.
The first set the tone. Tibo said the team was "locking in," and that the only things being worked on are "simplifications, more efficiency for more usage, groundbreaking features or new models." He added that "sometimes you have to invest ahead of the curve, but feedback is clear that you all want things to get simpler."
The second post, quoting the first, is the pledge itself: 28 days, one improvement or one full reset per day, "relevant for most codex/work users." Within hours it passed a million views, and the replies split almost evenly between cheering and sarcasm.
Three details matter more than the headline:
- "Relevant for most users." The bar is not "ship something," it is ship something that matters to the median Codex and Work user. A niche CLI flag does not count by the text of the pledge.
- "Or ship a full reset." A reset is the fallback, not a bonus. That implies the team expects some days to have no qualifying improvement.
- "Only things being worked on." This is a scope freeze. It reads as an internal admission that feature sprawl and unclear limits are the main user pain.
Why this landed on October 5
The pledge is a response to a specific month. Our coverage trail shows the pressure building:
- September 25 and 26: a Codex and Work outage, followed by a paid-account limits reset.
- Late September: a GPT-6 Astra quality postmortem and another reset.
- September 29 (DevDay): Dots, the always-on agent, shipped alongside Pro 500 and Ultrafast, while Pro 200 reopened with a smaller formula.
- October 1 to 3: GPT-6.1 Sol drew a massive load spike, and Tibo announced a global reset for all paid ChatGPT accounts.
- October 3: Claude Opus 5.5 took first place on the Epoch Capabilities Index, 0.84 points ahead of GPT-6 Astra, a narrow margin but a very quotable one.
- October 4: "6.1 coming soon" from Tibo, which we unpacked in what Tibo meant.
Read in order, this is a team that has been repeatedly paying down trust with resets. The 28-day pledge turns that habit into a policy, and into a public commitment to ship something every day.
The math: what could 28 days of this look like?
Nobody outside OpenAI knows the mix. These scenarios are illustrative arithmetic, not predictions, and assume a standard weekly Codex bucket that a global reset refreshes in full.
| Scenario | Resets in 28 days | Improvements in 28 days | Practical effect on your quota |
|---|---|---|---|
| Reset-heavy | ~20 | ~8 | Many effective fresh weeks; usage feels generous but the product barely moves |
| Balanced | ~14 | ~14 | Fresh bucket every other day on average, plus real feature progress |
| Improvement-heavy | ~4 | ~24 | Few resets; efficiency gains do the work, if the "more usage" promise is real |
Even the reset-heavy case is not unlimited use. Resets are one-time refills, not a higher ceiling. The deeper question is what happens on day 29, when the pledge ends and the underlying plan formula is what you are left with. That is why the "efficiency for more usage" half of Tibo's first post is the part to watch. Efficiency gains are the only item on his list that survives the 28 days.
Where the Oct 30 Pro 200 cut fits
The window starts October 5 and runs 28 days, which puts the last day around November 1 or 2. That means the October 30 change lands inside the pledge. From that date, Pro 200 Work and Codex usage drops from 20x to 10x the Plus allowance and GPT-6 Pro chat drops from 200 to 100 messages a week, per the subscriber email quoted in our Pro 200 breakdown. Existing seats keep old limits through October 29.
We have seen nothing from OpenAI that reverses this. Our read is that resets and efficiency wins will cushion the cut for a few days, not cancel it. If your Codex bill is the main reason you hold a $200 seat, price the post-October 30 plan, not the pre-cut one.
How to use your quota during the 28 days
The thread's best joke was that Tibo "turned usage into gambling: burn your Codex usage every day and pray it's a reset day." As a strategy, that is almost right, with one correction: do not waste usage, just do not save it.
- Stop hoarding. Earlier global resets replaced the remaining weekly balance. Saved quota is the quota most likely to evaporate.
- Front-load big jobs. Run long refactors and agent loops early in your window, then check Settings, then Usage each morning for a refresh.
- Log it. Keep a short daily record so you can tell, by day 28, whether the pledge changed your actual capacity or just your mood.
- Do not plan around announcements. There is no schedule. A reset that lands at 3 a.m. Pacific is useless to you at noon unless you check.
- Re-price before October 30. Compare the post-cut seat against API usage and Pro 500, as in our Sol pricing breakdown.
A simple log works:
| Date | What shipped | Reset? | Did it change my week? |
|---|---|---|---|
| Oct 6 | (fill in) | Y / N | Y / N |
| Oct 7 | (fill in) | Y / N | Y / N |
What "simplification" probably means (our inference)
OpenAI has not defined it, so treat the following as informed guesswork. Today a Codex user has to hold a lot in their head: GPT-6 Astra versus GPT-6.1 Sol versus Luna, a new Ultrafast speed tier, ChatGPT Work meters versus Codex meters, Pro 200 versus Pro 500, a first-month Dots allowance and a pile of overlapping features. One commenter in the thread suggested the cleanest simplification would be un-shipping something. That fits the "invest ahead of the curve" framing: remove, merge, and rename before adding.
The signal to watch is whether the first improvements are subtractive (fewer toggles, one clear usage meter) or additive (more modes). Subtractive changes would match what Tibo said users asked for.
How developers reacted
The reaction splits into four camps, and it is worth being honest that most of it is anecdote.
- Cautious optimism. Several developers said a daily commitment is better than silence, and that "all the best" is a fair response until the first week of results is in.
- Skepticism about gambling. The sharpest critique is structural: a reset is a refill, and tying quota to a daily lottery makes planning harder, not easier.
- Churn claims. Some posted that they had already moved a Pro subscription to Claude Max or moved API spend to Anthropic. Those are individual statements, not data, and similar threads appeared every time either vendor tightened limits.
- Hype that outran the facts. A widely shared reply claimed an overnight Claude Opus 5.5 run decompiled the Linux kernel into full source. It was posted as a joke-adjacent boast, we have seen no verification, and it should not be read as a benchmark.
One reply also argued the pledge reads like a "death rattle" unless a major model ships before day 28, a nod to the GPT-6.1 speculation Tibo himself sparked a day earlier. That is opinion, but it correctly identifies the real test: resets buy goodwill, shipped models and efficiency buy retention.
Should you move to Claude Code?
Short answer: not because of this thread, and not all at once.
The honest comparison has two sides. Claude Opus 5.5 does lead Epoch's index, and Claude Code remains the tool many developers prefer for long sessions and large codebases. But Anthropic has its own history with changing caps, which we tracked in the Claude usage limits timeline, and the older Claude Code vs Codex comparison showed the two tools trading advantages by workload: Claude for code quality and depth, Codex for speed and parallel agents.
The practical position is vendor-risk hedging:
- Keep both harnesses installed and authenticated.
- Route long, expensive refactors to whichever has headroom that day. The OpenAI Codex plugin for Claude Code lets you call one from the other.
- Compare cost per finished task, not per token.
- Re-evaluate on November 3, with 28 days of your own log in hand.
What would make this pledge succeed
A few honest checkpoints, in order of importance:
- A durable usage win. At least one efficiency improvement that raises how much work a fixed plan completes, not just a refill.
- A visible simplification. One fewer thing to configure, explain or pay for.
- A model or speed milestone. Whatever "6.1 coming soon" turns out to be, shipped on a clear date.
- A stable day 29. Limits that do not quietly tighten once the daily cadence ends.
If those four land, the pledge will look like a turning point. If it ends as 28 refills plus a Pro 200 cut, it will look like what its critics say it is.
Related reading
- Dots shipped, Pro 200 halved, Sol got a global reset
- What Tibo meant by "6.1 coming soon"
- Sept 25 to 26 Codex and Work outage reset
- GPT-6 Astra quality postmortem and reset
- GPT-6.1 Sol launch pricing and benchmarks
- Claude Opus 5.5 tops the Epoch Capabilities Index
- Claude usage limits timeline
- Claude Code vs Codex after the rate limit boost
Primary: Tibo Sottiaux's posts on X, October 4 to 5, 2026 · chatgpt.com/pricing
Details are accurate as of October 5, 2026, the day the pledge was posted. The scenario table is illustrative, social-media reactions are anecdotal, and OpenAI may change plan terms. Check Settings, then Usage, for your own limits.