TL;DR: Codex limits used to feel generous and workable, even for regular users like me who were not exhausting them. The recent reduction feels extremely severe, not minor. I understand the possible business logic behind pushing some users toward higher-tier plans, but this risks damaging long-term user habits and retention. In the AI market, losing users over preventable limit policy may be a much bigger mistake than it looks in the short term.
Codex limits used to reset quite often before their scheduled expiration. Because of that, I was not just happy with the product, but genuinely grateful. Also, except for maybe one time, I was never even a user who fully exhausted my limits. I was not someone using Codex at an extreme level and draining every bit of quota; in most weeks, I still had around 50% to 70% of my weekly limit left before it reset.
That is why I felt comfortable recommending Codex on social platforms. I often reassured people that they probably would not run into serious limit issues. Then, suddenly, the limits were tightened very heavily. This was not a small nerf. It was a major one. The difference between the old experience and the current rate of limit consumption is enormous.
I sometimes wonder whether the thinking is something like this: if one person upgrades to the $120 plan, that offsets the loss of six other users. That way, server load goes down while revenue stays balanced. I can understand that logic. However, there is something extremely important being overlooked here: user habits and long-term retention.
Right now, many of us may not leave immediately simply because of habit. But once the limits make regular usage feel impractical, and once that becomes noticeable enough, people may no longer have much choice but to look elsewhere. And once users move to other platforms, getting them to come back may be much harder than expected.
Especially at the dawn of the AI era, losing customers matters a lot. Losing them for a preventable reason like limit policy could become a much bigger mistake in the long run. On paper, losing six users and replacing them with one higher-paying user may look efficient. But if you eventually lose that one user too, then you have actually concentrated your risk. In other words, replacing six customers with one means a single future loss now costs six times more.
For that reason, I would strongly recommend reconsidering the recent changes to Codex limits. I believe there is room for a more reasonable balance between customer satisfaction and resource management.
