Monologue: What's Going On At Anthropic?

Why do things keep breaking? Used to be the coders could just launch claud and run a bunch of subagents to look productive. Think you can rely on claud code. Guess again, Dario Amede doesn't care about you, and he certainly doesn't care about your family. Anyway, to set the scene, I need to give you a few details about claud Code in general, which you don't pay for clawed code itself.

There's not like a CLAWD code subscription. When you subscribe to Anthropic service, you pay for a CLAWD subscription and you either pay twenty bucks, one hundred bucks, or two hundred bucks a month, with each tier having more access than the last, and access meaning right limits so you

can use it more. Though if you're wondering what that actually means, that's really by design, as all Anthropic tells you is that you've got a five hour limit and a weekly limit on them models, with a separate one for clawed Opus, which is their more expensive, more complex one. An important detail is that these subscriptions are also heavily subsidized. When you pay Anthropic two hundred dollars a month, you're not paying on a per token rate, which is what

AI starts have to do. They pay per million input and output tokens. No no, no, you just do stuff and stuff comes out, and you're able to do the same things you would I have used in the API. But you're burning tokens, and you can burn over two and a half thousand dollars in model tokens on a two hundred buck a month subscription, or at least you could, and some users have been able to spend upwards of

five thousand dollars on the same subscription rate. Limits are generally used to restrict people from burning as much, but the problem has become that for Anthropic to approach anything close to profitability would have to rate limit people into the well into the depths of Hell at this point.

Moving on. In the middle of March, Anthropics started a two week long promotional campaign where they doubled rate limits for off peak hours of using Claude, set to end on March twenty seven to twenty twenty six, a day before it was said to end on March twenty six to twenty twenty six, and Tropic would announce it was starting peak hours with claud code users maxing out their sessions faster than between the hours of five am and eleven pm Pacific Manday to Friday, with a spokesperson limply

adding that efficiency wins aren't named, of course, would offset this and only seven percent of users would hit the limits. All of this was sold as a result of managing the growing demand for Claude. Yes, spoiler alert, more than seven percent of users appear to have hit the limits and nobody seems to be feeling particularly efficient. Don't know where those gains are one use of one hundred dollars

a month. Max Plan complained about hitting sixty one percent of his session limit the five hour one after four prompts, which he found out based using a tool called cc Usage cost ten dollars and twenty six cents in tokens, he still spent ten percent of his subscription and four prompts.

Another said that they hit sixty three percent of the limit on their two hundred a month plan in the space of a day, and another hit ninety five percent off to twenty minutes of using their max plan, going to get one hundred dollars a month on that one.

Another person hit their max limit off to two or three things, don't know what they were, and another vow to cancel they're two hundred dollar a month subscription after hitting their weekly limit in the space of a day, saying that they and I'm going off of a translation from fucking Grock, so forgive me expected the premium experience between hundred dollars and what they got was constant limit stress. I've linked a lot of these in yesterday's free newsletter.

The SUBPRIMEI crisis is here. Really advise you read it, but also you can see how many people are mad, or just go on Twitter and search claud limits. It's it's not great now. While well, Anthropic technical staff member Lydia Halle posted that Anthropic was aware of people hitting usage limits in claud code way faster than expected, and that some investigation of some sort was taking place. It's hard to imagine that Anthropic had no idea that these limits were so severe, or that any of this was

a surprise. Now as I wrote this sentence that I'm reading on Tuesday, March thirty first, it doesn't appear that any changes have been made. People are still complaining about hitting their limits in a few prompts, and Anthropic is yet to update anyone. In my opinion, because these are the rate limits they decided were necessary to keep the

business going and roll their nasty ass into IPO. A few days previously, though, Anthropic could also accidentally, and I put that in quotation, Marks leaked the existence of their Cappy Boro and Mythos models to Fortune magazine, by which I mean they had a data cash to quote Fortune with over three thousand assets that was left open on

the internet and Fortune somehow found it. You know, Kanna reminds me of like a skeezy bloke dropping a Magnum condom out of his wallet in front of a woman and being like, hey, yeah, you see that, or whoops, whoops. In any case, this massive leak also included absolutely fucking nothing about the models themselves, other than that they're a step change better and that their cybersecurity features were so very scary that they would have to roll them out slowly.

I'm I don't know, man, I no insult to the people at Fortune. I just don't think they're on fucking Showdan looking up aws buckets. So wonder how this got there? Now? I will say, at first I was completely sure that this was a deliberate leak, because three thousand assets left open on the Internet and none of them have any info on the model, but just scary things like ooh with too much better. But now I'm not so sure, because yesterday Anthropic accidentally leaked the entire source code of

its claud code coding interface. The leak was a result of a reference in claud Code's MPM package, the thing that you queried download clord code onto your computer via the terminal that led right back to a ZIP archive on Anthropic servers per the register that contained the source code.

I will add that there's an ongoing discussion about what actually caused this problem with someone hacker news, hacker news, how can use I'm just going to keep it saying it might be a problem with bun, the packaging tool used to allow people to download clawed code that Anthropic acquired in December of last year. To be clear, this isn't a leak of Anthropics models, but it's still an unbelievably large leak, one that exposed claud code's innards to

the entire Internet and all of their competitors. And while I imagine using the source code is illegal on some level, I can't imagine there's any reason their competitors can't take a look or that they're you think that they're all sitting around being like, oh, I absolutely can't, I mustn't. It's not okay, especially when this is a company that fucked over just about anyone building anything that you build on top of a claud code subscription, thinking about how

they treated open code. By the way, anyway, now is a great time to remind you that claud code create a Boris Churney said at the end of December that one hundred percent of his contributions to clawd code were written in claud code, and told Lenny Richitski in February twenty twenty six the coding is now solved for most use cases. I assume that the use cases that haven't been solved include making sure that there isn't a direct link from the claud code installer to a cloud Flare

storage bucket with its source code ready to download. That one just doesn't seem like they got that one pinned down. I've also had that. The same interview added that Cherney hadn't written a single line of code since November twenty twenty five, which is I mean, those of you playing at home, I know many of you really enjoy my game show is that good? And I just have to ask you is that good? Anyway? I feel like I very recently warned everybody about the very obvious dangers of

allowing lms to write all of your code. These models do not have thoughts or knowledge, or really anything other than probabilistic generations of outputs based on training data and a lot of it, and they're quite complex. Nevertheless, this means that any time you choose to just accept the code they generate without reading it thoroughly, you're choosing to trust thing inherently untrusted, whether it doesn't think or have knowledge.

This is just the beginning of us finding out the ugly cost of software engineers trusting large language models to write their code at scale. Llms are good at writing lots of code, which in turn means that the code requires far more time to review, which assumes you do review it, which becomes far more difficult when you're constantly pressured to ship more and more and more software every

fucking week. Or if you're Boris Journey and you believe that shipping software fast is the same thing is shipping good software. I'm now thinking about the launch of anthropics claud cowork, a product that can allegedly manage files on your computer or draft documents or some such Billshere nobody seems to be able to explain it. At the time, Boris Journey proudly boasted that clawed cowork was built in around a week and a half, which makes perfect sense, says.

Almost immediately, a story came out of a guy who tried to reorganize his wife's desktop files using cowork, which is one of the literal things on the cowork website. It tells you to do, only to watch it delete. Every single one of his folk goes permanently. He was only able to recover them thanks to a nicloud backup. This is the future that these so called large language

model companies want for you. Bad software chipped quickly hyped by a captured media, doesn't give a damn about whether the services are the things you pay for or functional or useful, and doesn't even bother using the tools or understanding them. And nobody else appears to be discussing how inherently deceptive these companies have. Become a subscriber to Claude who paid for an annual subscription in December twenty twenty five now as a product that has significantly less value

thanks to egregious rate limits. And that's also like if you can even do your job anymore, I imagine people in the twenty buck a month subscription are really fucking suffering. And in general, anthropics models seem to go and look at the Reddit if you want to understand more. They seem to oscillate in utility and effixi based on the time of day you use them and their proximity to

a new model launch. This is a really weird story that I've never been able to get to the bottom of, but there have been reports for like over a year of different mode from Anthropic from OpenAI from whomever getting dumber at random times. No one's able to get to the bottom of it. I've heard people saying they quantize the models make them smaller during the day. I don't know if you have any information about this, shoot me at shoot me a piss on clerk, or email me

a easy at better offline dot com. That's Echo Zeta at better offline dot com. What we do know is that claud Code, one of the few popular AI products, is built with the same disregard for safety and customer happiness as the rest of Anthropics, astonishingly shitty business that burns billions of dollars with no end in sight. Chief claud Code slopagandist Boris Cherney doesn't give enough of a fuck about his customers to actually read his code, and I imagine the same goes for a lot of other

engineers within the company a big tech at large. The consequences are already obvious. In the last few months, A coding tools brought down Aws twice and lost Amazon hundreds of thousands of orders and led to a security breach inside Meta mere weeks ago. And now how these tools of leaked claud code source code. How much is enough to make people wake up to the inherent dangers of using these models the right software at scale. I guess

we're gonna find out. I'll see you on Friday for another monologue.