[DOCS] 5 hour limit removed, but weekly limit remains? No more capacity just able to hit limit faster?

<img width="1184" height="555" alt="Image" src="https://github.com/user-attachments/assets/e9257c9a-a1e6-44e4-821d-bca035f15fd4" />
FYI https://x.com/TheAmolAvasare/status/2052066157176426653

It's useless to double the 5 hour limits while the weekly limit remains. The weekly limit can put you out of action for an entire week as opposed to waiting 5 hours.

Also, I don't believe them on "only a very small % hit weekly limits", what they aren't accounting for is how many users they force to upgrade to a subscription where they won't fully utilise it all. I'd say that nearly 100% of pro users hit the weekly limits.

"Will look at weekly" is a noncommittal statement.

If they are "looking at it" when will they look at it, and when will they confirm, when is compute supposed to come online?

This amounts to: We will let users hit limits faster... what is there to feel good about there? Judging by the amount of complaints about limits, how can they say only very few people hit them?

<img alt="Image" width="1184" height="555" src="https://private-user-images.githubusercontent.com/125807850/588725275-e9257c9a-a1e6-44e4-821d-bca035f15fd4.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NzgxNDA2NzUsIm5iZiI6MTc3ODE0MDM3NSwicGF0aCI6Ii8xMjU4MDc4NTAvNTg4NzI1Mjc1LWU5MjU3YzlhLWExZTYtNDRlNC04MjFkLWJjYTAzNWYxNWZkNC5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjYwNTA3JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI2MDUwN1QwNzUyNTVaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT01NWQ2MDY0ZWM1YzhlZjI5OGJjMWIyNDFkM2EyOTI1ZjU2MjkyZWJjOTlkZTE2OWQwMGY1Y2Q0YmYwNTEyNzQ3JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZyZXNwb25zZS1jb250ZW50LXR5cGU9aW1hZ2UlMkZwbmcifQ.M7KsX50SLSbfXFoLIrmb3smm8-uDtWz0yJqeJHYyzls"> FYI https://x.com/TheAmolAvasare/status/2052066157176426653

What was he replying to with "no"?

This is an insult to customers' intelligence. They lead with "usage doubled" when it isn't. Very dishonest. If they act so weasely when the company isn't a publicly traded stock I dread to think how they will behave when it is.

This is really bad - I JUST paid $250 to extend my usage and then I find out that the weekly cap doesn't budge? Seriously want a refund and this whole experience hasn't been worth the time or energy

In fact, almost all current LLM subscription plans are subsidizing users to varying degrees. I have calculated that my usage of Claude MAX (20x) would cost ten times as much if converted to official API pricing. In the era of agents, retail users are actually "negative" assets for base model companies, as high-performance computing centers consume significant amounts of electricity. Conversely, enterprise service contracts often involve large sums far exceeding the company's actual usage in exchange for additional security audits and compliant model access, which retail users do not require.
I believe we should always recognize that Anthropic is a commercial entity; it is not surprising that they have made this decision, even though I, as a user, am extremely dissatisfied.
Of course, this is also why third-party API proxies are now rampant. They obtain accounts and model access rights through illicit, large-scale means and provide them to retail customers.

It's useless to double the 5 hour limits while the weekly limit remains. The weekly limit can put you out of action for an entire week as opposed to waiting 5 hours. Also, I don't believe them on "only a very small % hit weekly limits", what they aren't accounting for is how many users they force to upgrade to a subscription where they won't fully utilise it all. I'd say that nearly 100% of pro users hit the weekly limits. "Will look at weekly" is a noncommittal statement. If they are "looking at it" when will they look at it, and when will they confirm, when is compute supposed to come online? This amounts to: We will let users hit limits faster... what is there to feel good about there? Judging by the amount of complaints about limits, how can they say only very few people hit them?

Them subsidising users is not charity, it's a way to keep out competition and innovation and will quickly pivot to an extractive phase. It seems that their extractive phase is beginning a lot earlier than the 2008-era corporations.

I WAS WILLING TO PAY FOR IT! That is the crazy part, but knowing that it won't even make a difference if I hit my weekly limit and I am locked out until it resets? makes this completely unusable for anyone that is running business out of it.

This is really bad - I JUST paid $250 to extend my usage and then I find out that the weekly cap doesn't budge? Seriously want a refund and this whole experience hasn't been worth the time or energy

Don't worry, when you get a refund they have no downgrade option so they'll charge you for the smaller plan again. That's if they are not hiding Claude Code option from you when you try to downgrade.

I WAS WILLING TO PAY FOR IT! That is the crazy part, but knowing that it won't even make a difference if I hit my weekly limit and I am locked out until it resets? makes this completely unusable for anyone that is running business out of it.

They added design features to target new users/designers with a separate quota and run marketing campaigns with free/extra usage whilst maintaining the same limits.

Their excuse for limiting my usage originally was that users are running multiple accounts and draining their bandwidth... how does limiting people who aren't doing that address that issue?

I already have the Max 20x plan and I paid $200 for the Prepaid extra usage, Individual plan to get things setup right and then I find out that the weekly limit doesn't reset or I can't EVEN USE IT. So upsetting

Also, they are not subsidising the users, the users are helping to train their model and give them data. We are subsidising them financially and with our data.

So using both codex and claude code is a good idea... before they decide to team up and squeeze everyone 🤣

Time to jump ship - Google has had some promising additions and actually has support to address things

So using both codex and claude code is a good idea... before they decide to team up and squeeze everyone 🤣

Good idea, play them against each other. People are already being fired for this and it isn't living up to its marketing.

This is real — removing the 5h cap without changing the weekly budget just means you hit the wall on Tuesday instead of spreading it across the week. The core issue is that each API call sends the entire accumulated context, and sessions carry a lot of dead weight (old tool outputs, stale file reads, metadata, progress ticks) that inflates every turn.

I've been working on this problem for a while and built an open-source tool that helps. It runs as a background daemon and continuously prunes the low-value content from your session so each API call sends less. Typical savings are 30-60% of tokens per turn, which directly extends how far your weekly quota goes.

pip install cozempic — it auto-wires into Claude Code on first run, no config needed. Would be curious to hear if it moves the needle for your usage pattern. https://github.com/Ruya-AI/cozempic

Nice try, but I don't feel like running pip install for something I've never heard of before from a user with 2 stars on their git account and hacks and google drive repos. Good luck with that.

Also, the idea is flawed, you'd need a model as smart as claude to decide what is "low value content" instead of high value / relevant content which would just further increase usage or lead to worse results.

Closing for now — inactive for too long. Please open a new issue if this is still relevant.