Description: Tibo's 28-day Codex challenge offers a daily product improvement or a full usage reset. On Day 9, a desktop storage fix met the same familiar request: “Reset.” Here's why the giveaway has overshadowed the update.
Tibo announced a change to how the ChatGPT and Codex desktop apps save their state to disk. The expected result: roughly 60% less stored data.
One reply to the October 13, 2026 US-time announcement had a much shorter feature request: “Reset.”
The odd part is that both sides have a point. Freeing disk space matters when your computer is running out of room. But if your AI allowance is exhausted, an empty drive is not the thing stopping you from shipping work.
So what is Tibo's 28-day challenge?
Starting October 5, the premise was simple: for 28 days, deliver one clear improvement relevant to Codex and ChatGPT Work users, or give users a full usage-limit reset.
Day 8 was a reset. Day 9 was a storage improvement.
Those rewards work on different clocks. A product fix should make future sessions better. A reset can let an eligible user continue today. When somebody has hit their weekly ceiling halfway through a task, the second option has unusually direct value.
The challenge accidentally turned release notes into a daily “What did we get?” draw.
The fix was real. It just fixed a different problem.
Tibo's 60% figure concerns the amount of application state saved on disk. It's a projected reduction from a change to the desktop apps' storage approach, not a measured promise that every user's entire project directory will shrink by 60%.
Old test artifacts and build folders sitting inside a software project are a different cleanup problem. So is the usage meter in the account dashboard.
Someone short on storage can celebrate. Someone short on quota can appreciate the work and still ask for a refill. Someone facing both problems has an excellent argument for wanting both.
Why “give us another reset” can be more than a joke
Consider one heavy user's reported pattern: Codex hits its limit after roughly 12 hours of sustained use, whereas Claude remains usable for about four days under that person's workflow.
Four days is 96 hours; 96 divided by 12 is eight. On that simplified example, matching eight Codex-sized periods would mean the original allowance plus seven additional full refills.
That's an illustration of one person's experience, not an official eightfold difference between the services. Both providers say usage depends on the model, task, settings, and session or weekly windows. The user may also prefer one model's output quality, but that's a judgment about a particular workload.
What people feel they are buying isn't a headline benchmark score. It's the number of real tasks they can finish before the meter shuts them down.
The free refill becomes the expectation
There's a predictable comedy to repeated surprises.
First reset: “Wow, thanks!”
Next reset: “Nice, another one.”
Then a new update lands: “Cool. Is there a reset with that?”
In Japan, this mood invited a deliberately shameless nickname: “reset beggars.” Think of the queue for a free refill that has somehow become more popular than the café itself.
Officially, a promotional reset does not permanently raise the plan's allowance, and future resets aren't promised. A saved reset that you activate later also differs from an automatic global one.
That's precisely why the joke lands. The bonus helps, but it doesn't erase the original frustration with usage limits.
Tibo set out to ship something useful each day. The most devoted users have developed an extremely efficient review process:
“Great fix. Now, about that reset?”
Sources: Tibo's Day 9 announcement · 28-day reset record · OpenAI usage guidance · Claude usage guidance
