The picture that makes me say “yay” much less typically than “god rattling it”.
Subscription-based coding brokers akin to Claude Code and Codex famously have 5-hour and weekly quotas on their LLM utilization. Each of those are comprehensible: 5-hour quotas assist stagger utilization so the servers don’t get overloaded, and weekly resets forestall customers from dumping a complete month’s price of utilization right into a single day which a) additionally prevents overload and b) stops the person from simply unsubscribing after they accomplish that. Each Anthropic and OpenAI have performed round with quota limits, from doubling them for a restricted time to even eradicating the 5-hour quota.
These mannequin suppliers can reset the weekly quota for all customers, typically gifted as compensation within the occasion of technical glitches on their finish. If, for instance, you’ve got a $100/mo Codex plan, then a weekly reset is price $25 to you assuming you absolutely devour your quota—these days with the price of top-tier LLMs like Fable 5 and GPT-5.6 Sol, that’s simpler to do. Nonetheless, these quota resets aren’t telegraphed and are usually not introduced by company-owned channels: until you comply with particular folks akin to Thibault Sottiaux (Tibo), the engineering lead for Codex, you simply take a look at your weekly quota and see it’s at 100%. You possibly can even get a quota reset hours earlier than your precise reset and never profit from all of it.
Not too long ago, within the wake of the discharge of Fable 5/GPT-5.6, Anthropic and OpenAI have been doing weekly quota resets for his or her harnesses way more continuously. Up to now two weeks, OpenAI has instantly reset the Codex weekly quota six occasions: July 9, July 10, July 10 (once more), July 14, July 15, and July 17. That’s not even moving into the not often mentioned banked reset system for Codex, the place OpenAI gave quota resets July 12 and July 13 which will be manually used at any time however expire inside 30 days.
https://codex-resets.com tracks Codex resets.
Nobody needs to be the weirdo who complains about actually getting free stuff as a result of if the quota resets cease, they would be the one blamed for it. Regardless of that, I’m going to be the weirdo who complains about actually getting free stuff.
As an individual who doesn’t like losing cash if I can simply keep away from it, I attempt to use as a lot of my weekly quota as I can. I used to be on the $20/mo Codex plan however with the promise of GPT-5.6 on the horizon and with the frequency I stored hitting the 5-hour limits, I upgraded to the $100/mo plan. Through the use of GPT-5.5, I used the 5x immediate capability to construct and take a look at various bold initiatives, however that’s a subject for one more weblog publish. I used to be in a position to constantly exhaust my weekly quota over the complete week, though I needed to have a browser window open with my Codex utilization to continuously monitor it. I’ve set reminders on my cellphone for the precise time a 5-hour or weekly quota resets so I can preserve working extra prompts—as an apart, I want there was a canonical platform endpoint to programmatically test Codex utilization quantity and the quota reset occasions so I may simply vibecode an app to handle for me. Earlier than the discharge of GPT-5.6, I assumed “ought to I intentionally exhaust my quota all on the weaker GPT-5.5 and gamble that OpenAI does a quota reset to greet GPT-5.6?” (I didn’t exhaust my quota and OpenAI did certainly do a quota reset)
GPT-5.6 Sol is certainly a terrific mannequin that does reside as much as the hype, and I’ve needed to create new initiatives simply to have an excuse to check its limits. The speed of quota utilization is about the identical as GPT-5.5 even with a immediate working always. Regardless of that, nearly each time a reset has occurred after the GPT-5.6 launch, it has been when my weekly quota has been at 50+%, which makes me really feel like I “wasted” $12. There’s additionally the issue that when the quota resets happen, the weekly reset time is unset, so I’ve to enter some immediate simply to set off it once more even when I’m doing one thing else.
@emollick / X
After the flurry of quota resets over the previous two weeks, the supposed pleasure has as an alternative change into annoyance as I now need to urgently create new concepts to spin down the quota earlier than it inevitably resets once more. Random rewards are supposed to present a dopamine hit however I find yourself with a web dopamine deficit from each the sense of wasted quota and having to abruptly change my plans. I admit that this may occasionally simply be an indication I’m burnt out and have to take a break.
Tibo polls if there are too many resets, with just a few thousand responses.
Resets of the weekly quota for all customers should be ludicrously costly for these corporations, though once you already spend billions of {dollars} in CapEx per yr it’s probably a rounding error. The current surge of quota resets is probably going not a coincidence: July has been an insane month in LLM releases, with not simply Fable 5 and GPT-5.6 Sol pushing frontier fashions even additional, but in addition Grok 4.5, Muse Spark 1.1, and Kimi K3 providing extra choices throughout the price/utility curve. The cynical take is that weekly quota resets aren’t supposed to be enjoyable serendipity, however as an alternative supposed to stop energy customers from experimenting with sufficiently aggressive rivals as soon as the quota naturally runs out.
I don’t count on weekly quota resets to final endlessly even when competitors intensifies, as a result of if quotas preserve resetting this continuously they gained’t matter in any respect. It will as an alternative give me an incentive to downgrade from the $100/mo plan again to the $20/mo plan to keep away from losing quota, which I don’t suppose is OpenAI’s supposed aim.
