Considering all the problems they've been having with over-charging Claude Code users over the past few weeks it's the very least they could do. Max subscribers are hitting their 5 hour usage limits in 30-40 minutes with a single instance doing light work, while Anthropic have no support or contact mechanism for users that they respond to.
impulser_ 22 hours ago [-]
"Max subscribers are hitting their 5 hour usage limits in 30-40 minutes with a single instance doing light work"
This has not been my experience at all. The only time I even got close to this is multiple long sessions that had multiple compacts.
The key is if you hit compact, start a new session.
namelosw 18 hours ago [-]
This hasn't been my experience either. I personally find the max plan is very generous for day-to-day usage. And I don't even use compact manually.
However, when I tried out the SuperPower skill and had multiple agents working on several projects at the same time, it did hit the 5-hour usage limit. But SuperPower hasn't been very useful for me and wastes a lot of tokens. When you want to trade longer running time for high token consumption, you only get a marginal increase in performance.
So people, if you are finding yourself using up tokens too quickly, you probably want to check your skills or MCPs etc.
bicepjai 6 hours ago [-]
As a regular user, I hit these walls so often. I am experimenting with local model and open code. I am hoping to see some good results with qwen3 coder
mcast 20 hours ago [-]
It's known that Anthropic's $20 Pro subscription is a gateway plan to their $100 Max subscription, since you'll easily burn your token rate on a single prompt or two. Meanwhile, I've had ample usage testing out Codex on the basic $20 ChatGPT Plus plan without a problem.
As for Anthropic's $100 Max subscription, it's almost always better to start new sessions for tasks since a long conversation will burn your 5-hour usage limit with just a few prompts (assuming they read many files). It's also best to start planning first with Claude, providing line numbers and exact file paths prior, and drilling down the requirements before you start any implementation.
deaux 20 hours ago [-]
> It's known that Anthropic's $20 Pro subscription is a gateway plan to their $100 Max subscription, since you'll easily burn your token rate on a single prompt or two.
I genuinely have no idea what people mean when I read this kind of thing. Are you abusing the word "prompt" to mean "conversation"? Or are you providing a huge prompt that is meant to spawn 10 subagents and write multiple new full-stack features in one go?
For most users, the $20 Pro subscription, when used with Opus, does not hit the 5-hour limit on "a single prompt or two", i.e. 1-2 user messages.
pastel8739 15 hours ago [-]
Today I literally gave Claude a single prompt, asking it to make a plan to implement a relatively simple feature that spanned a couple
different codebases. It churned for a long time, I asked a couple very simple
follow up questions, and then I was out of tokens. I do not consider myself to be any kind of power user at all.
deaux 14 hours ago [-]
The only time I've ever seen this happen is when you give it a massive codebase, without any meaningful CLAUDE.md to help make sense of it and no explicitly @ mentioning of files/folders to guide, and then ask it for something with huge cross-cutting.
> spanned a couple different codebases
There you go.
If you're looking to prevent this issue I really recommend you set up a number of AGENTS.md files, at least top-level and potentially nested ones for huge, sprawling subfolders. As well as @ mentioning the most relevant 2-3 things, even if it's folder level rather than file.
Not just for Claude, it greatly increases speed and reduces context rot for any model if they have to search less and more quickly understand where things live and how they work together.
visarga 12 hours ago [-]
I have a tool that scans all code files in a repo and prints the symbols (AST based), it makes orienting around easy, it can be scoped to a file or folder.
fragmede 8 hours ago [-]
> spanned a couple different codebases
It's either that, or you have a lot of skills loaded or something. I use Claude for hours a day and usually don't run out of tokens.
mcast 7 hours ago [-]
I should note this only happens in Claude Code, not the web UI. Since CC is agentic and spawns subagents on prompts that require a lot of thinking.
It will spend a lot of time grokking the codebase, which would consume more tokens on larger projects.
Foobar8568 16 hours ago [-]
I am on $100 max subscription, and I rarely hit the limit, I used to but not anymore, but then again, I stopped building two products at the same time and concentrate to finish up the first/"easiest" one.
8cvor6j844qw_d6 9 hours ago [-]
I'm considering dropping the Max plan for the API.
Using the Max plan with tools like OpenClaw violates Anthropic's ToS [1].
The API gives you the same flexibility without the risk of getting your account suspended.
At ten times the price. At least you keep the ability for them to charge you a hundred a month.
runako 19 hours ago [-]
> you'll easily burn your token rate on a single prompt or two
My experience has been that I can usually work for a few hours before hitting a rate limit on the $20 subscription. My work time does not frequently overlap with core business hours in PDT, however. I wonder whether there is an aspect of this that is based on real-time dynamic usage.
cantalopes 14 hours ago [-]
i never had these issues with gemini cli using google vertex endpoint, and i never even reached $50 per month
i don't want to think about how to hack a tool i'm paying for not locking me out because "i promped wrong"
darqis 3 hours ago [-]
It's my experience however.
daliusd 9 hours ago [-]
I wonder what do you mean by "if you hit compact". Claude Code does not show used tokens.
ben_w 9 hours ago [-]
When I used it before Christmas (free trial), it very visibly paused for a bit every so often, telling me that it was compressing/summarising its too-full context window.
I forget the exact phrasing, but it was impossible to miss unless you'd put everything in the equivalent of a Ralph loop and gone AFK or put the terminal in the background for extended periods.
bavell 6 hours ago [-]
Run /usage or configure your statusline
novaleaf 8 hours ago [-]
if you enable verbose mode, it does.
However I run like 3x concurrent sessions that do multiple compacts throughout, for like 8hrs/day, and I go through a 20x subscription in about 1/2 week. So I'm extremely skeptical of these negative claims.
Edit: However I stay on top of my prompting efficiency, maybe doing some incredibly wasteful task is... wasteful?
coffeefirst 1 days ago [-]
Wait, if it burns through a $200/mo allotment that fast, what are all these people who “barely wrote code anymore” doing?
input_sh 1 days ago [-]
You hit a vague, never-quite-explained 5h window limit that has nothing to do with what you're doing, but with what every user is doing together. It's totally not downtime, you're just "using it too much" and they're telling you to fuck off until the overall usage slows down.
The order of priority is: everyone using the API (you don't want to calculate the price) → everyone on a $200/month plan → everyone on a $20/month plan → every free user.
00deadbeef 20 hours ago [-]
Yeah it's way too vague.
This morning: (new chat) 42 seconds of thinking, 20 lines of code changed in 4 files = 5% usage
Last night: 25 minutes of thinking, 150 lines of code generated in 10 new files = 7% usage
exclipy 18 hours ago [-]
I think the first message consumes disproportionately much percentage because it's not cached and includes the system prompt, tools, etc.
input_sh 15 hours ago [-]
Here comes the "you're using it wrong" defence!
Let's be perfectly clear: if user actions had anything to do with hitting these limits, the limits would be prominently displayed within the tool itself, you'd be able to watch it change in real time, and you'd be able to pinpoint your usage per each conversation and per each message within that conversation.
The fact that you cannot do that is not because they can't be bothered to add such a feature, but because they want to be able to tweak those numbers on the backend while still having plausible deniability and being able to blame it on the user.
Instead, the little "usage stats" they give you is grouped by the hour and only split between input and output tokens, telling you nothing.
kaizenb 14 hours ago [-]
Interesting. What do you think the reason for not being transparent on this matter?
input_sh 12 hours ago [-]
For the same reason they use "tokens" instead of kilobytes: so that you don't do the conversion yourself and realise that for example spending a million "tokens" on claude-opus-4.6 costs you anywhere from $10 (input tokens) to $37.5 (output tokens). Now, 1 million tokens sounds pretty big and "unreachable" until you realise that's about 4 megabytes of text. It's less than three floppy disks of data going back and forth.
Now let's assume you want to send a CD worth of data to Opus 4.6. 700 megabytes * $10 (price per million input tokens) / 4 (rounding down one megabyte to roughly 250k "tokens") = $1750. For Opus 4.6 to return a CD amount of data back to you: $37.50 * 700 / 4 = ~$6.5k.
A terabyte worth of data with a 50:50 input/output ratio would cost you $5.7 million. A terabyte worth of data with a 50:50 input/output ratio on gpt-5.2-pro would cost you $25.2 million. (Note: OpenAI's API pricing still hasn't been updated to reflect 5.3 prices.)
So we get layers upon layers upon layers upon layers upon layers of obfuscation to hide those numbers from you when you simply subscribe for a fixed monthly fee!
nl 12 hours ago [-]
Do people care about how many bytes they are sending or receiving?
Most people care about getting the right bytes.
kaizenb 10 hours ago [-]
Oh man. Thanks for explaining. That sounds like a dark pattern.
potamic 14 hours ago [-]
Blurring the cost-benefit analysis in the interest of downplaying the costs.
inglor_cz 8 hours ago [-]
I definitely noticed that Claude is faster on Central European mornings, e.g. deep night in the US.
8note 22 hours ago [-]
you can just watch the limit on the claude usage settings view.
itd be nice to know how much the session context window applies wrt token caching, but disabling all those skills and stopping sending a screenshot every couple messages gets that 5hour limit and weekly limit a bunch better
huflungdung 24 hours ago [-]
[dead]
smcleod 22 hours ago [-]
It hasn't always done this, it's a relatively recent problem in the last 1-4 weeks (roughly). (note: I'm on the $160AUD/mo plan, so I think that's $100USD).
heavyset_go 20 hours ago [-]
> what are all these people who “barely wrote code anymore” doing?
Writing self-serving LinkedIn productivity porn
1 days ago [-]
SOLAR_FIELDS 20 hours ago [-]
“Light work” is a pretty bold statement my dude. I run max for 8+ hour coding sessions with 3-4 windows where I’m babysitting and watching the thing and I never even get session warnings. The only time I bump up against limits is on token hungry tasks like reverse engineering 3M+ LOC codebases or 5-6 agents generating unit tests in parallel. Something tells me that what you call “light work” is not remotely the same as what I consider “light work”
cantalopes 14 hours ago [-]
exactly my experience. i am on pro subscription and when coding with claude in console i can only do about 30 min of work before i have to wait for 4 hours. i actually code more by chatting with claude via poe.com than using the subscription that i have
brulard 12 hours ago [-]
pro is the $20, right? It runs out quickly, especially using opus. But what do you expect for that kind of money? For serious work at least Max $100 is needed.
mercutio2 1 days ago [-]
It hadn’t occurred to me this was a billing bug.
That would be heartening, if I wasn’t consuming tokens 10x as fast as expected, and they just had attribution bugs.
Do you have references to this being documented as the actual issue, or is this just speculation?
I want to support Anthropic, but with the Codex desktop app *so much better* than Anthropic’s combined with the old “5 back and forths with Opus and your quota is gone”, it’s hard to see going back
smcleod 22 hours ago [-]
Yeah I think it's either a billing bug, or some sort of inbuilt background sub-agent loop gone wild inside Claude Code, if you have a look at recent issues on the Github relating to 'limits', 'usage', 'tokens' you'll see a lot of discussion about it: https://github.com/anthropics/claude-code/issues?q=sort%3Aup...
klipklop 20 hours ago [-]
Yeah I can work for minutes in the $20 Claude plan but potentially hours with Codex. It’s just not worth bothering.
alpineman 13 hours ago [-]
Wasn't customer service going to be one of the first things to be fully automated by AI? :D
chaostheory 19 hours ago [-]
Experienced this with the $20 sub, but not with Max (yet).
subscribed 15 hours ago [-]
Yeah, I find very capable but won't paying them exactly because they cheat /steal like this.
CosmicShadow 1 days ago [-]
Is this one of those, Hey turn off the overcharge protection because you can go $50 into debt for free, and then maybe you'll just keep going and not notice you owe us an extra $500 type of situations?
Ethee 1 days ago [-]
I was worried about this when I turned it on myself, but under the usage panel it shows that it limited my spending to just the $50 and that auto-reload is off, so it doesn't seem this would be the case.
brulard 12 hours ago [-]
I turned on this overspend and limited the spending to $20. A day later I checked my spending, I had used "295%" of my limit. Almost $60. No idea why it didn't respect my setting.
nezhar 1 days ago [-]
There is also an Auto-reload option, I think it is off by default.
dude250711 1 days ago [-]
You guys are putting a lot of trust into vibe-coded software...
sxg 1 days ago [-]
The underreported story is how we over-trust human-coded software.
falloutx 24 hours ago [-]
Too true, lets not trust any software.
ninjha 21 hours ago [-]
most slop is human slop, who do you think claude learned from :)
22 hours ago [-]
JumpCrisscross 20 hours ago [-]
> You guys are putting a lot of trust into vibe-coded software...
Nope. I'm putting a lot of trust in American Express and the continued availability of Claude competitors.
Doesn't appear to include the new model though, only the state-of-yesterdays-art (literally yesterdays).
teaearlgraycold 20 hours ago [-]
I have access to 5.3 in the Codex desktop app. I'm really not pleased with the desktop software (insane performance issues) but the LLM is good.
Reagan_Ridley 24 hours ago [-]
I'll pass on this $50, but please hire real human and fix your crappy app, Claude!
This bug has been for years: in Claude (web or app), if you create a new chat at the middle of existing chat thinking or tool calling, the existing chat will be broken, either losing data, or become unusable.
It's unbelievable Anthropic worth hundreds of billions but can't fix this.
moffkalast 24 hours ago [-]
My favorite bug is when I spend 5 minutes writing something and send it, then something breaks and refreshes, deleting the whole thing. Like, thanks, much appreciated, please waste more of my time.
sobjornstad 23 hours ago [-]
Literally every time if I press the "stop" button while it's writing back for the first conversation turn because I notice I forgot something and want to correct it, my prompt is lost.
Anytime I run into a bug like this, part of me wants to go calculate how much of humanity's collective time has been wasted by one company not fixing a trivial bug. It's got to be a lot.
jimmydoe 23 hours ago [-]
Yes!! The whole chat part of app is flaky as hell!
Sometimes I think it’s better to just use code tab to chat knowing that’s more reliable.
bonsai_spool 24 hours ago [-]
Sometimes you can recover it by refreshing the page where you lost the text.
nmfisher 19 hours ago [-]
I get this regularly too, it's very annoying. I've started using one git worktree per Claude session specifically to avoid this issue.
nezhar 1 days ago [-]
You can use this if you started your Pro or Max subscription before Wednesday, February 4, 2026 at 11:59 PM PT.
I received €42, top up was not required and auto-reload is off.
VikingCoder 9 hours ago [-]
It'll be nice to see how quickly I burn through this free money with auto-reload off.
Unless they somehow say, "You owe us $181,902, bucko! Turning off auto-reload just meant 'Don't add more money to pay off your debt!' It didn't mean, 'Don't incur more debt!' SUCKER!"
darqis 3 hours ago [-]
I'll certainly not enable extra usage. This is a trap.
Keep your "credits"
chinathrow 16 hours ago [-]
I think this enabled a default overusage allowance of USD 50 on my account - which I'm sure was never active before.
onion2k 1 days ago [-]
You started your Pro or Max subscription before Wednesday, February 4, 2026 at 11:59 PM PT.
Ah well. Back to Codex.
dr_dshiv 1 days ago [-]
They are trying to breed more whales
23 hours ago [-]
2001zhaozhao 1 days ago [-]
I mean, it's a nice gesture. I use extra usage a little bit when the quota runs out and I'm still in the middle of a task. (Afterwards I switch to other agents that track limits monthly instead of by 5 hour window.) This $50 credits should last a while.
magospietato 24 hours ago [-]
Expires after 60 days per the link. In case you're thinking of stretching it out beyond a couple of months.
CyberDildonics 21 hours ago [-]
The AI spam on this site is way out of control. It's every single article on the front page.
xvector 16 hours ago [-]
So weird seeing comments like this on HN. AI is the most revolutionary technology in programming and computing, perhaps since the first programs were made. Of course it's going to be the most talked about topic.
CyberDildonics 10 hours ago [-]
I remember all the other "most revolutionary technologies" too. It isn't just the most talked about topic, almost every single link is some point release or random blog entry about "my journey".
jmkni 1 days ago [-]
So much cynicism here, I for one am loving Opus and appreciate this
copilot_king 11 hours ago [-]
[dead]
Rendered at 22:42:29 GMT+0000 (Coordinated Universal Time) with Vercel.
This has not been my experience at all. The only time I even got close to this is multiple long sessions that had multiple compacts.
The key is if you hit compact, start a new session.
However, when I tried out the SuperPower skill and had multiple agents working on several projects at the same time, it did hit the 5-hour usage limit. But SuperPower hasn't been very useful for me and wastes a lot of tokens. When you want to trade longer running time for high token consumption, you only get a marginal increase in performance.
So people, if you are finding yourself using up tokens too quickly, you probably want to check your skills or MCPs etc.
As for Anthropic's $100 Max subscription, it's almost always better to start new sessions for tasks since a long conversation will burn your 5-hour usage limit with just a few prompts (assuming they read many files). It's also best to start planning first with Claude, providing line numbers and exact file paths prior, and drilling down the requirements before you start any implementation.
I genuinely have no idea what people mean when I read this kind of thing. Are you abusing the word "prompt" to mean "conversation"? Or are you providing a huge prompt that is meant to spawn 10 subagents and write multiple new full-stack features in one go?
For most users, the $20 Pro subscription, when used with Opus, does not hit the 5-hour limit on "a single prompt or two", i.e. 1-2 user messages.
> spanned a couple different codebases
There you go.
If you're looking to prevent this issue I really recommend you set up a number of AGENTS.md files, at least top-level and potentially nested ones for huge, sprawling subfolders. As well as @ mentioning the most relevant 2-3 things, even if it's folder level rather than file.
Not just for Claude, it greatly increases speed and reduces context rot for any model if they have to search less and more quickly understand where things live and how they work together.
It's either that, or you have a lot of skills loaded or something. I use Claude for hours a day and usually don't run out of tokens.
It will spend a lot of time grokking the codebase, which would consume more tokens on larger projects.
Using the Max plan with tools like OpenClaw violates Anthropic's ToS [1].
The API gives you the same flexibility without the risk of getting your account suspended.
[1] https://www.anthropic.com/legal/consumer-terms (Section 3-7)
My experience has been that I can usually work for a few hours before hitting a rate limit on the $20 subscription. My work time does not frequently overlap with core business hours in PDT, however. I wonder whether there is an aspect of this that is based on real-time dynamic usage.
i don't want to think about how to hack a tool i'm paying for not locking me out because "i promped wrong"
I forget the exact phrasing, but it was impossible to miss unless you'd put everything in the equivalent of a Ralph loop and gone AFK or put the terminal in the background for extended periods.
However I run like 3x concurrent sessions that do multiple compacts throughout, for like 8hrs/day, and I go through a 20x subscription in about 1/2 week. So I'm extremely skeptical of these negative claims.
Edit: However I stay on top of my prompting efficiency, maybe doing some incredibly wasteful task is... wasteful?
The order of priority is: everyone using the API (you don't want to calculate the price) → everyone on a $200/month plan → everyone on a $20/month plan → every free user.
This morning: (new chat) 42 seconds of thinking, 20 lines of code changed in 4 files = 5% usage
Last night: 25 minutes of thinking, 150 lines of code generated in 10 new files = 7% usage
Let's be perfectly clear: if user actions had anything to do with hitting these limits, the limits would be prominently displayed within the tool itself, you'd be able to watch it change in real time, and you'd be able to pinpoint your usage per each conversation and per each message within that conversation.
The fact that you cannot do that is not because they can't be bothered to add such a feature, but because they want to be able to tweak those numbers on the backend while still having plausible deniability and being able to blame it on the user.
Instead, the little "usage stats" they give you is grouped by the hour and only split between input and output tokens, telling you nothing.
Now let's assume you want to send a CD worth of data to Opus 4.6. 700 megabytes * $10 (price per million input tokens) / 4 (rounding down one megabyte to roughly 250k "tokens") = $1750. For Opus 4.6 to return a CD amount of data back to you: $37.50 * 700 / 4 = ~$6.5k.
A terabyte worth of data with a 50:50 input/output ratio would cost you $5.7 million. A terabyte worth of data with a 50:50 input/output ratio on gpt-5.2-pro would cost you $25.2 million. (Note: OpenAI's API pricing still hasn't been updated to reflect 5.3 prices.)
So we get layers upon layers upon layers upon layers upon layers of obfuscation to hide those numbers from you when you simply subscribe for a fixed monthly fee!
Most people care about getting the right bytes.
itd be nice to know how much the session context window applies wrt token caching, but disabling all those skills and stopping sending a screenshot every couple messages gets that 5hour limit and weekly limit a bunch better
Writing self-serving LinkedIn productivity porn
That would be heartening, if I wasn’t consuming tokens 10x as fast as expected, and they just had attribution bugs.
Do you have references to this being documented as the actual issue, or is this just speculation?
I want to support Anthropic, but with the Codex desktop app *so much better* than Anthropic’s combined with the old “5 back and forths with Opus and your quota is gone”, it’s hard to see going back
Nope. I'm putting a lot of trust in American Express and the continued availability of Claude competitors.
Doesn't appear to include the new model though, only the state-of-yesterdays-art (literally yesterdays).
This bug has been for years: in Claude (web or app), if you create a new chat at the middle of existing chat thinking or tool calling, the existing chat will be broken, either losing data, or become unusable.
It's unbelievable Anthropic worth hundreds of billions but can't fix this.
Anytime I run into a bug like this, part of me wants to go calculate how much of humanity's collective time has been wasted by one company not fixing a trivial bug. It's got to be a lot.
Sometimes I think it’s better to just use code tab to chat knowing that’s more reliable.
Go to https://claude.ai/settings/usage, turn on extra usage and enable the promo from the notification afterwards.
I received €42, top up was not required and auto-reload is off.
Unless they somehow say, "You owe us $181,902, bucko! Turning off auto-reload just meant 'Don't add more money to pay off your debt!' It didn't mean, 'Don't incur more debt!' SUCKER!"
Ah well. Back to Codex.