@_mauriciorubio @d4m1n You're both retards. Someone reverse-engineered the binary and found 2 caching bugs
English
VR
4.4K posts
VR
@ValueRaider
Amateur quant trader. Improving OSS Python toolkit mostly yfinance
Rotterdam, The Netherlands Tham gia Haziran 2022
36 Đang theo dõi92 Người theo dõi
@d4m1n I can confirm it’s not a bug
Mauricio Rubio 🎧@_mauriciorubio
Funny how you think this is a “bug” and not deliberate behaviour. If you’re business is a slot machine based on consumption of tokens, doesn’t really feel that your interests would be aligned with reducing or eliminating token consumption. If anything, the contrary. Something they would never publicly admit of course. But let’s get real. KPIs and business models drive behaviour. And remember, in a casino filled witj slots machines, you can “feel” like you’re winning sometimes, sometimes, but in the end, the house always wins.
English
prediction:
Anthropic will come out saying this was a bug (it wasn't lol)
+ roll back all the changes & reset limits because this is the brand assassination of the century
Dan ⚡️@d4m1n
how did Anthropic drop the ball like this this is my entire feed rn 💀
English
@sagemmyx6 @MerrynSW That's a lie. They just disrupted infrastructure. I never saw them build infrastructure.
English
Is there a small silver lining watching the just stop oil people understanding what happens if you do indeed just stop oil.
English
Alex Volkov@altryne
PSA: If you've been running out of Claude session quotas on Max tier, you're not alone. Read this. Some insane Redditor reverse engineered the Claude binaries with MITM to find 2 bugs that could have caused cache-invalidation. Tokens that aren't cached are 10x-20x more expensive and are killing your quota. If you're using your API keys with Claude this is even worse. This is also likely why this isn't uniform, while over 500 folks replied to me and said "me too", many (including me) didn't see this issue. There are 2 issues that are compounded here (per Redditor, I haven't independently confirmed this) : 1s bug he found is a string replacement bug in bun that invalidates cache. Apparently this has to do with the custom @bunjavascript binary that ships with standalone Claude CLI. The workaround there is to use Claude with `npx @anthropic-ai/claude-code` 2nd bug is worse, he claims that --resume always breaks cache. And there doesn't seem to be a workaround there, except pinning to a very old version (that will miss on tons of features) This bug is also documented on Github and confirmed by other folks. I won't entertain the conspiracy theories there that Anthropic "chooses" to ignore these bugs because it gets them more $$$, they are actively benefiting from everyone hitting as much cached tokens as possible, so this is absolutely a great find and it does align with my thoughts earlier. The very sudden spike in reporting for this, the non-uniform nature (some folks are completely fine, some folks are hitting quotas after saying "hey") definitely points to a bug. cc @trq212 @bcherny @_catwu for visibility in case this helps all of us.
English
@TheSpeculator0 probably hard to fix fast cause it’s a compute capacity issue - and they were “reasonable” in their forecast and purchasing
English
Can anthropic stop shipping 100 new peripheral features per day and just make sure claude code works
thanks
English
@mitsuhiko Armin you are gaslighting. Claude Code is a vibe-coded hot mess. No testing, no proper version control.
x.com/altryne/status…
Alex Volkov@altryne
PSA: If you've been running out of Claude session quotas on Max tier, you're not alone. Read this. Some insane Redditor reverse engineered the Claude binaries with MITM to find 2 bugs that could have caused cache-invalidation. Tokens that aren't cached are 10x-20x more expensive and are killing your quota. If you're using your API keys with Claude this is even worse. This is also likely why this isn't uniform, while over 500 folks replied to me and said "me too", many (including me) didn't see this issue. There are 2 issues that are compounded here (per Redditor, I haven't independently confirmed this) : 1s bug he found is a string replacement bug in bun that invalidates cache. Apparently this has to do with the custom @bunjavascript binary that ships with standalone Claude CLI. The workaround there is to use Claude with `npx @anthropic-ai/claude-code` 2nd bug is worse, he claims that --resume always breaks cache. And there doesn't seem to be a workaround there, except pinning to a very old version (that will miss on tons of features) This bug is also documented on Github and confirmed by other folks. I won't entertain the conspiracy theories there that Anthropic "chooses" to ignore these bugs because it gets them more $$$, they are actively benefiting from everyone hitting as much cached tokens as possible, so this is absolutely a great find and it does align with my thoughts earlier. The very sudden spike in reporting for this, the non-uniform nature (some folks are completely fine, some folks are hitting quotas after saying "hey") definitely points to a bug. cc @trq212 @bcherny @_catwu for visibility in case this helps all of us.
English
I wonder how much of the recent complaints about Opus are just because people now take advantage of the much larger context window (where the dumb zone now starts at around 6% of context :P
English
@Credit_Junk Tbf, threatening invasion for negotiation only works if it actually looks legit.
English
*TRUMP SAYS DEAL WILL 'PROBABLY' BE REACHED
He says this while massing troops and equipment at the border, including special forces and amphibious landing craft.
English
@RBowles44666 @nummanali Because it doesn't implement agents properly? They should specify what model to run with, not everything needs to be Opus.
English
@nummanali I thought everyone was already using then until I see a bunch of posts about skills to create skills.
Wildly inefficient and reinventing the wheel.
English
The upgraded Skill Creator from Claude Code team has built in eval framework that runs tests with child agents to determine effectiveness and drive improvement
What you have in the skill is years of researcher experience encoded
Very surprised by effectiveness
github.com/anthropics/ski…
English
@DonMiami3 Might be the aim for America to send ground troops to Yemen and not Iran. Iran is a Fortress. Trump might be satisfied with a “win” over Yemen with his tail between his legs. Won’t solve his Hormaz issue but I think he knows he can’t win there.
English
Few are prepared if the Red Sea closes.
The course of humanity may change.
FinancialJuice@financialjuice
⚠ BREAKING: Pentagon prepares for weeks of ground operations in Iran - Washington Post
English
@jacobLadder9 @GBNEWS "Stop blocking license" costs nothing, we can still invest in renewables. Let private money risk their own money in drilling.
What is complicated about this?
English
@GBNEWS It would not take a penny off people's bills.
The licences given out over the past 14 years have extracted very little.
Energy security means investing in renewables.
English
@hector_drummond Probably includes per-capita consumption of all business and industry in Cambridgeshire.
English
Does the average Brit really use 137 litres of water a day? Or are there more people in Britain than we think?
Simon French@Frencheconomics
Rationing: Water thetimes.com/life-style/pro…
English
The surreal thing about this for me is that I don't even understand what their argument was in their own terms.
Lyn Alden@LynAldenContact
Still can’t believe they did this.
English
@shaunralston @extliqprovider @Stable @2 Yup, massive improvement, although ~24 hours had passed since last session so not impossible something else changed.
At least try it.
English
English
Anthropic cutting Claude Opus 4.6 rate limits with 0 warning is frustrating 🤬 my $200/month MAX plan rate limited in 2 hours, mid-project. never had that happen with Codex. Any DEVs thinking of dropping MAX but keeping Pro, for ad-hoc use?
English
@extliqprovider @shaunralston Downgrade client to @stable or @2.1.45. Something broke with the 1M update.
English
@Osinttechnical Why are they parking tankers out in the open at a base that has previously been hit?
English
Iranian forces appear to have damaged or destroyed multiple USAF KC-135 tankers parked in the open at Prince Sultan Airbase in Saudi Arabia.
English
@kayintveen @GrahamJCampbell But it wasn't a subsidisation change. There was/is a bug in client, because downgrading fixes it.
English
@GrahamJCampbell tbh exactly right. the people who rage-cancelled were already on the free tier mentally. real usage patterns never change over a pricing shift
English
Hot take. Everyone cancelling their Claude subscription because the subsidisation is now less were shit customers who were never going to spend sufficient money anyway. Google learned this the hard way attracting the bottom of the market.
English
@AlexGlaucon @andrew_lilico You're still being a retard.
Government should not stimulate demand for a commodity *in a shortage*. You keep ignoring that part, why?
English
@ValueRaider @andrew_lilico That’s a fair point. So - stepping back from the edge of regardless - should the government avoid stimulating demand for things involving atoms, that it also pays for. Eg refuse to engage in wars, or try to manage down the population so it doesn’t have to support education
English
But *why*?? *Why* should there be any support package at all? Our policy community has gone mad. Stark raving mad.
Steven Swinford@Steven_Swinford
Exclusive from @benclatworthy Rachel Reeves is preparing to announce a limited package of support for energy bills over the summer and is instead focused on a more significant bailout in the winter The chancellor believes support for households who use gas for heating can wait until the autumn. The energy bill price cap, set by Ofgem, the regulator, is expected to jump by about £330 in July, forecasts suggest. That means average household bills will hit £1,972, according to Cornwall Insight, a consultancy. Treasury officials said Reeves was expected to allow this to go ahead. She believes the picture will be clearer by autumn and any support package should be introduced when gas consumption jumps with the arrival of colder months Government figures show that households use 78 per cent of their annual gas consumption during the winter thetimes.com/uk/politics/ar…
English
@ValueRaider @andrew_lilico Would you also say governments shouldn’t build roads as that just drives up energy consumption too? I’m trying to get to the underlying principle
English
@altryne @AnthropicAI Nope.
200max plan user here.
I get throttled so the free users / lowest tier can get served?
English
First official response from @AnthropicAI to the sessionGate thing (where-in folks on pro/max plan are hitting quota much faster than before)
Over 500 folks commented earlier this week to my thread saying this happened to them.
Are you guys satisfied with this response?
Thariq@trq212
To manage growing demand for Claude we're adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged. During weekdays between 5am–11am PT / 1pm–7pm GMT, you'll move through your 5-hour session limits faster than before.
English