Elevated error rate across multiple models
status.claude.com
status.claude.com
Because then you can install it without depending on a package manager?
Understand that 99% are comfortable trusting downloads. They know that it's just as easy to sneak backdoors into source code as it is to sneak backdoors into executables.
See also: XZ hack.
I would never runa script without reviewing it. I would install a package from a distros repository without reviewing the contents, however, because I can trust that a distro maintainer has reviewed it, that anyone else in the community can review it, and that that the bytes I'm downloading are the specific bytes I'm supposed to be downloading.
If you run a script off the open internet, you're being massively irresponsible. There are so many attack vectors that could be used here, and they are much easier to implement than something like the massive social engineering attack that was XZ.
My experience does not match yours.
Well, I suppose that matches the "move fast and break things" credo. lol.
I personally try and use my operating systems package manager for all applications (in this instance, dnf on Fedora). The moment you start adding a million repos, third party package managers, Flatpaks, Snaps, random curl install scripts, etc, it becomes extremely unmaintainable.
What happens if the curl application depends on openssl, maybe with a legacy or specific cryptographic function? I assume the curl script will either install the required version, or include the relevant libraries right? Now that is outside of the system package managers scope, meaning updating openSSL to avoid some vuln now requires extra work.
What happens when you go from Fedora 44 to 45? You should be checking all your applications are supported on whatever version or operating system you are running. There is a decently high chance you run into dependency issues when some lib version is updated.
Package Managers are (generally) more secure and simple to use for an end user (they are using the OS to begin with).
Curl scrips are easier for the software developers.
And it's not just sometimes - it's most people.
So. They have their option too. And it's at the top, where most people can find it.
"You can mine and refine your own iron ore so you can have a metal fork without depending on modern infrastructure."
Sure, but why.
But to answer your question directly - I suspect they simply don't care, and just take the easiest way.
curl -fsSL https://claude.ai/install.sh | bash
Oh wait (from another comment under this article): > https://pi.dev/models is throwing an internal server error for me.
(pi.sh also documents other install methods, like `npm`, on their homepage)
If trust and security is the issue, unfortunately "better" ideas like hashpipe [1] never achieved critical mass
[1] https://news.ycombinator.com/item?id=9318286Indeed, plenty of these scripts often act as a "what OS and packager do we have" mux. Just look at the source of this one, for example.
When you support an open source project at scale and/or with less savvy users, you come to see the benefit of "here, just f'ing slam this into your shell and we'll figure it out" installers. I know I have.
However they all look the same to the end user.
That's a feature and also a potential source of problems since users cannot tell if that particular application they want to install Is implementing the installer correctly or not. The outcome is that most users just trust that application (possibly because it's popular and trusted) and that's fine but it also trains the public that this installation method is ok and that gives a positive feedback for other applications to also offer their software using that installer pattern until at least one of such packages is implemented very badly or sneakily malicious.
If only a curl had a flag where you pass the sha256 of the file and it first checks it against the buffered file before outputting it to stdout.
That would singlehandedly resolve this whole kerfuffle.
The install instructions will be a slightly longer one liner and that's fine because people copy paste it anyway
Package managers: ecosystem is fragmented, requiring a long list of distro- and package-manager-specific instructions. Many scripts already install through package managers, they simply make the user’s life easier.
Flatpaks: These are clearly designed for desktop applications, with CLIs treated as an afterthought. They may be the best long-term hope, but today they are definitely not as convenient or widely available as a simple script.
If you care about adoption, `curl | sh` is the only real option today, which is why virtually all project show it as the first option.
There's plenty of big projects that don't suggest you curl a script right into your shell.
If you have curl, you're probably on Linux. Just use the package manager like an adult.
Users need to figure out which of the 10+ package managers they should be using, then run several commands. If something fails, the error messages are often cryptic and not easily configurable by the distributor.
And that’s before getting into the many rough edges of package managers. Most of them flat-out refuse to handle configuration and leave that part to the end user. Now you also need to document how to edit YAML and restart a systemd service. With an install script this is also solved.
For power users, this always looks trivial. In practice it raises the barrier to entry and can meaningfully affect adoption if your product is often used by less technical people.
IMO a lot of the blame falls onto the package manager ecosystem refusing to take into account very valid needs and claiming they aren't real / desirable.
In what world does a user have to choose between 10 package managers? Each distro has exactly one. There are also only about three, maybe four main package managers out there.
A shell script being piped into bash has so many more ways to break than a package. And if yhe theory is that package managers are fickle (they aren't), then how does adding more complexity help?
It is much simpler, much safer, and easier to maintain a package than an install.sh, eapecially for a big project.
Configuration can be handled by a script, yes. Here's a crazy idea: Your package can include scripts for configuring the software. It's almost as if most packages do. The scripts/utilities could even restart a systemd service for you.
Unless you're talking about configuring your build, in which case we're dealing with an experienced developer who will have no trouble just cloning the repo and building from source.
My biggest issue is: if we're dealing with someone who can't use a package manager, we're dealing with someone who doesn't have the capacity to judge how safe a script downloaded off the internet is. This does not drive linux adoption, it drives botnet adoption.
I'll stop there. I'm not interested in continuing this discussion when it's being conducted in bad faith.
Rather than argue with those of us who are pointing out messy realities, this commenter might be better served filing a bug against any number of the projects that offer installation this way, asking them to remove it, and see if it lands any better.
Technical purity/superiority isn’t the only factor, or even the most important one, driving projects to offer quick installers like this.
I would also appreciate it if you actually talk about something concrete rather than simply claiming to be right. You shouldn't pipe stuff from the internet into your shell.
Are you claiming that's about some highfallutin "technical purity"? Is it technical purity to check inside the bag when you buy a pig in a poke? No, that's common sense. It's common sense to have some degree of knowledge about what programs you execute on your computer. As root, at that.
Nobody would argue that it's categorically safe/good/smart to blindly pipe a script into your shell; and for the record, I agree. I would also readily agree that habituating users to doing this probably creates new, more general risks especially among how less-technical users interact with their CLI.
However, the realities of the "real world" make it popular for a reason, in light of those negatives; tons of scaled projects continue to offer a 1-liner. So we have to ask, why? They'd probably say that's because it (a) improves project adoption, and (b) reduces "install broken" tickets.
You have to address the non-technical merits and goals to get behavior to change here, and sadly, I don't think anyone has done that.
But who cares about me? I'm not currently maintaining one of these (though I did once). My suggestion to bring your argument to an active project was genuine: try it! I'd be delighted to see you bring about the change you want.
[PS: The commenter I replied to originally used the term "bad faith", which they've since edited]
I am arguing in good faith about the merits of the approaches. I am engaging with the points of argument being brought up from the opposing side of the argument (see above). I am not veering off on side-tracks, unlike you, for example.
There' a simple good faith argument (that I have been making) which you can try responding to: Running arbitrary code from the internet without checking is bad. There is some effort needed to package software, but that is not that much effort in the grand scheme.
what has worked over time is having computers of various types in schools, where teachers teach students and let them play with it.
nobody teaches about the command line, so nobody knows what to do with it. its also inscrutible without a useable help view, unless you already know how to use the terminal
I’m all for higher Linux adoption on desktop, but there’s still a lot of resistance to making less-skilled users the primary target instead of power users.
Teaching can help, but if it takes 50 hours to learn the basics of Linux versus 5 hours for Windows, it’s a losing battle.
It's plain horrible. You could have, for example, a compromised server serving malware but only one out of every 100 download. The only signature you rely on is TLS.
Proper package distribution are using proper signatures schemes, are decentralized, even for some offer reproducible builds (meaning you can rebuild the whole package yourself and verify your build matches), etc.
Hashpipe is an attempt at reproducing some of those guarantees. Not unlike container pining using hashes. It at least fixes the "Jack and John installed this already and I know I'm getting the same version as they did".
Proper software distribution is signed, reproducible and ideally also uses some proof-of-existence for the hashes.
My bet is this: in the face of the countless supply chain attacks, we'll see more and more people getting very serious about security, including the security of software distribution. And curl bash'ing won't be part of it.
Is it likely? No. Can it happen? Yea.
Just make it `curl -o <file> <url> && sh <file>` and this entire problem is gone.
(At least for bash scripts, I’m not sure whether these are POSIX syntax to be frank.)
Maybe security should be at a higher position on our priority list.
The careless days are ultimately over but we still don’t act like that.
oh wait...
"curl -fsSL https://claude.ai/install.sh | bash"
(right from https://claude.com/product/claude-code)
Further - what the flicking fuck do you think an installer is going to do on your system? Not run any commands? Because I've written installers for every platform... they ALL can run commands.
So what exactly is the complaint in this comment? If you want to go read the install script - knock yourself out (or hell, point your agent at it...).
You think it's hard to obfuscate shell calls from inside a built executable?
What it tells us is that you're probably searching for reasons to grouse about AI.
/s
It's just people who have internalized "don't paste commands from the Internet into your terminal" and aren't thinking about exactly what makes pasting commands from the Internet into your terminal dangerous, and how that applies to this specific case.
One caveat is that it doesn't do MCP tools, but can wire them up with bash (or use CLIs if those are available).
As far as I can tell, they tick the same boxes- but one has the support of a big boy model provider.
If someone wanted to fork Codex and make a community-maintained version that supports third party models, that would be great, because I liked Codex better than OpenCode for the most part.
Maybe you've found workarounds. Maybe you're using an old version of Codex. Maybe you have your own soft fork. I don't know. But I used to be able to use Codex with self-hosted models, and I gave up on that about a month ago as they kept breaking that.
From watching Pr's and issues- seems like openai at least wants to come across as if theyre supporting non-oai models :/
You would think they would support their own GPT-OSS model, but, not really anymore. I wish they would release a GPT-OSS 2, but this doesn't fill me with confidence.
OpenCode is nice if you don't want to do a lot of research and just want to get started right away. The OpenCode Go plan for $5 a month for your first month is a great way to do this, with good models to choose from and reasonable usage limits for a beginner.
I occasionally use OpenCode.
I try to use Codex and Antigravity as much as I can, often using it as a secondary agent (due to different usage pricing models than API). The same skills and MCPs work across harnesses.
Edit: I don’t use Claude Code simply because I already have enough to deal with and don’t see a major advantage to their harness. I use Opus credits from my Google subscription on the rare occasion I need them.
Cursor is also worth checking out particularly at the $20 a month price tier. If you have Grok you effectively already have it too.
I expect to have a completely different answer a year from now. The main “lift” we’ve gotten from AI tools is our clients now get an Android + iOS app + macOS app + Electron + PWA to go with whatever web based app they want us to build, at essentially the same original price. (There’s also a CLI and a TUI, but so far none of them care about that…)
We just made the decision to start adding MCPs to apps. Gonna be an interesting conversation in a few weeks when I can tell my business contact he can use his favourite chatbot to now plug in directly to the custom app he bought from me.
So I'm sold on that level alone. Good stuff.
Try Zed[1] for GUI and pool[2] for TUI.
[1] https://zed.dev/
I asked a simple question about my codebase. Pi: >60s and 100k tokens, I cancelled and gave up Claude (pointed at the same deepseek model): 17s 15k tokens
I burned my $5 GLM budget in 15 minutes.
So idk what the Pi hype is about cuz this is my 3rd disappointing try
Not worth it in my opinion.
You'll get 10x more value by using MiMo 2.5 or DeepSeek Flash which is what I use for most code.
I get the power of LLMs, and I do find them useful. But I find them useful in much the same way I find a really good set of random tables useful, or a good set of rules for procedurally generating something like a star sector for a science fiction campaign.
For my day job developing software, and for the RPG campaigns and books I run and publish today, LLMs are, in many cases, random tables on steroids. After using them for two years, even with all their improvements, I am continually reminded by the results I get that, at the heart of it, I am still dealing with what amounts to randomly generated content.
Yes, I know it is more accurate to call the process probabilistic rather than random. And yes, somebody can construct a technically deterministic setup with fixed weights, fixed seeds, fixed sampling parameters, and a frozen runtime environment. But that is like saying you can recreate a rainstorm if you get a thousand butterflies to flap their wings in exactly the right way. It may be technically true, but it is not how the technology behaves in normal day-to-day use.
For practical purposes, given the same prompt and the same apparent starting conditions, the result can differ each time you use a model. The outputs will often be highly correlated, and often useful, but they are not deterministic software in the ordinary sense.
So far, I am failing to see how the inherent probabilistic nature of the technology can be fully overcome. I understand how we got to where we are today from older neural net technology, including the systems used for vision and sound. What we have now can be very useful. But my view is that it is being badly oversold and overhyped. Its probabilistic nature is being vastly underestimated, and that is a major reason for much of the weirdness and many of the failures we keep seeing.
In tabletop roleplaying, there have been times when hobbyists relied too much on procedurally generated content and ultimately got burned by it, either through campaigns that were not as fun or products that were subpar. Each time, the lesson was the same: there is no substitute for human judgment.
Any workflow or technology incorporating LLMs has to keep humans in the loop, and not merely as rubber stamps. The human has to remain the primary decision maker.
Those chemical interactions and quantum effects lead to emergent properties like judgment, experience, context, accountability, and an understanding of consequences. Those are not properties that LLMs possess, regardless of how useful their output can be.
That is not to say that, in the future, LLMs won’t be used as part of other systems that add some of those properties. But that is not what we have today, or what can be seen in the foreseeable near future.
What makes you say that? Other than the usual "I'm a human, and humans must be very special, so when something that's not a human does X, it's either not real X, or X wasn't important in the first place".
It highlights, in my eyes, that "critical flaws" of LLMs are the same exact flaws that humans routinely suffer from. Sometimes LLMs have it worse, but sometimes they have it better too.
LLMs do improve release to release though. Humans are more of a mixed bag.
This is common in image generation pipelines because if you find an image you really like, you can store the seed and then reproduce it with small tweaks, otherwise - to quote Borges - “Look at it well. You will never see it again" User-facing deterministic pipelines do exist for generative AI.
I know you make this argument in your post, but that's really the answer if you want repeatable results. For a classifier or a detector, determinism is a requirement, but for an LLM non-determinism desirable property because it feels like a more natural conversation. The downside is it's extremely difficult to replicate a response without pointing the model to an earlier conversation.
And specifically for the RPG case, don't you want non-determinism? You don't want the model spinning up the same identical person if you say "Generate me an NPC character sheet for an innkeeper". This was a complaint that people had in the past, that models would regurgitate the same scenarios or the same jokes.
Where I suspect DMs run into trouble is not randomness, but lack of self-consistency in worldbuilding. Say you generate an NPC and then refer back to them later and the model gets some details wrong. You could compare to a system like Dwarf Fortress where everything down to the genealogy and faction relationships are rigidly generated.
LLMs can contribute quite reliably given very narrow prompts and short horizons (keeping turns low and context brief). If you chain a bunch of these narrow contributions together and define guardrails (structured outputs, online evals, other-llm-as-judge/jury, etc...) you can produce a very repeatable workflow that reliably delivers to defined service levels.
The obvious issue being - you've got to define the workflow and implement all the guardrails, not hope that the LLM will infer them during a session or a one-shot prompt.
For many of these problems, I think it is likely that no deterministic algorithm can exist because the problems are fundamentally underspecified. E.g. a common task in computer vision is generating a 3D depth map from a 2D image. This is inverting a lossy projection, so any solution must be a least partially a hallucination.
I think we just have to accept this. It's a different type of algorithm, built out of statistics instead of logic, with different strengths and weaknesses compared to traditional software.
Now, biology exists. It’s wet and messy and impossible to understand (we haven’t invented the microscope yet). That doesn’t mean biological study is not worth doing.
I deeply hope we never reach the point where that’s overcome. What we’ve seen over the past few years is how AI will destroy humanness from pretty much the entire digital realm. It’s by far the most evil, anti-human technology ever created, corrupting everything it touches. The last thing we need is for it to become reliable
Perhaps it would be better if it actually worked.
Computed from the page’s own data for 2026-03-26 through 2026-06-23:
- Partial outage: 43h 15m 1s
- Major outage: 6h 46m 48s
- Total affected time: 50h 1m 49s
- Major-only uptime: 99.6861%
These stupid SLAs in the SaaS contracts should be reframed like this: 99.9% during working hours, not 99.9% overall. Would also give the SaaS vendor leeway to only guarantee 90% availability outside working hours, and then do their maintenance tasks in those windows.
> Optimize your bottom line for token spending so you collect $$$$
> Release Ultracode feature that optimize for Token Spending (a.k.a Dynamic Workflows)
> Tokenmaxxing achieved + 529 Overloaded unsustainable APIs everywhere
This video, wow: https://www.threads.com/@founder__growth/post/DZz_9Ikj3Wx
Out of desperation, I moved to ChatGPT and it's working better than I remember. All these companies are playing games under load, under failure. No wonder we can't agree on what's good for what.
ClaudeCode still has a 99.27 % uptime
ClaudeCowork has 99.52 % uptime
ClaudeForGovernment has 99.93 % uptime
API Error: 529 Overloaded. This is a server-side issue, usually temporary — try again in a moment. If it persists, check https://status.claude.com.
[*] https://claude.ai/api/organizations/<ORG_ID>/chat_conversations/<CONV_ID>/completion
"it's look like when the lights turning off, we return to socialize lol"
Today is the Latvian holiday of Jāņi, to mark the passage of the summer solstice: https://en.wikipedia.org/wiki/J%C4%81%C5%86i
Grab yourselves some beer or beverage of choice and some cheese (we usually have caraway cheese), alongside skewered meat and get some rest!
I mean, what else am I going to do while Claude is down, write code manually, like they did in the 90s or something?
I don’t prompt Claude anymore. I have loops running that prompt Claude and figuring out what to do. My job is to write loops.
— Boris Cherny, head of Claude Code
Except now it's the "AI did it" fallacy where if you know a company uses AI, even infra scaling issues must be due to AI, and if you had just used less or no AI, you would have been spared even though that has never been true.
The usual response to this goes something like "well they made claims that AI is good" therefore anything short of perfection supposedly debunks the claim.
This is literally they saying they are letting their LLM run wild(ish) and seeing the status.claude.com we can see the result.
This is a case where the outcome is the direct result of the engineering practices like the ones they describe.
PS: Yes I use Claude, Coded, Amp and Cursor agents every day so I am not saying here LLMs are not valuable.
LE: They did not made claims that "AI is good" they made claims that developers/computer engineers are not needed anymore in the near future. Thats is a stronger claim and has a direct relation with a product they have which needs computer engineering (yes infra counts too) and which seems to be down more than we expect as a good quality bar.
Unless you have inside knowledge of their infra ops and management tools, it is just guessing and blaming veganism. For all we know it could be tools from Nvidia or anyone else failing under massive load.
It could be the veganism. Some things are. Leaping to it as the only possible explanation for every ailment is exactly the fallacy.
I have not stated anything. I just replied to a metaphor which is not needed cause here we talk about engineering problems handled by engineers in a tech company. I give you something else where this line of thought could be wrong: culture beats (and destroys) engineering practices unless regulated by law. In this case yes this is not because of LLMs but because of company culture.
Still hard to know where the line draws because Anthropic talks about solving computer science for good as in humans need not apply.
Could be a hardware issue, could be a datacenter issue. Could be anything. So it could also be a software issue right?
This is not ideology. This is talking about root causes and I replied to someone that started saying this is not because of them promoting and using LLMs to the maximum. Could be it is not because of that. But it could be because of LLMs.
Keeping a company accountable when they try to sell a service that will replace engineers is not ideology. Ideology will be to not use any company that uses LLMs. But pointing out the disconnect between the public discourse and the status.claude.com is a simple idea.
Can you tell me that all those red lines there are infra?
The issue is that it's a thought-terminating cliche, and it would be nice to have one place on the internet that isn't just who can post one the fastest with the most glee to the giddy seal-clapping of the audience.
So not sure what we are debating here: I see first hand companies jumping full on using LLM for _everything_ for the last 6 months (of course Anthropic longer) and without guardrails and good engineering practices the number of incidents, downtime is increasing.
Look at status.claude.com - Anthropic could at any point come out and say all those are due to third party providers.
I am also not saying here Anthropic is worse than other scaleups. But they do something different: they come in front of us and tell us they have better engineering practices.
Why can't it be simply the case that Anthropic is struggling by their own accord? Infra scaling isn't a solved problem, much less with new, complicated, ever-changing, stateful LLM requests.
Pretty much every API-service-centric company I've worked at was in some constant state of either triaging or thinking about infrastructure health, often due to the familiar cascading problems of a necessarily distributed system.
But now with the AI scapegoat, we rewrite history to pretend us humans solved infra scaling, so any issues today must be caused by AI and any related superstitions we want to tack on.
They can and it is normal. I have said it is normal for scaleups specifically at a similarity growth rate.
What we (or at least I) critique here is coming out in the world and announcing that coding is done while having a product that has a status page full with red stripes. Yes, could be infrastructure, could be third party integrations could be a lot. But a lot of what is there is software. And yes, some parts is hardware. Unless the root cause is culture. In that case as I mentioned in another comment there I give them that: LLMs cannot solve culture.
Again the difference here is that the other scale-ups with similar _scaling_ issues are not talking about how we should all just use LLMs for everything and that learning to code is not required anymore.
So I am not saying the real issue is not infra or integration with third parties. What I am pointing at is: "don't talk that you don't need engineering while you - yourself - have engineering problems that need engineering solutions and still have not solve them".
Also you are getting out of your way to brainstorm possible root causes that will let them get away with this cognitive dissonance (or is there a better them in communication). Let them do the explanation and defend their position as they are the ones attacking the computer science engineering.
> Let them do the explanation and defend their position as they are the ones attacking the computer science engineering.
This once again boils back down into: because they make claims about LLMs being good, I get to make any claim I want, and if if they didn't want me to make my claim, they shouldn't have made theirs.
It seems reactionary rather than earnest.
You've accused me and someone else of "brainstorming" reasons why they might have infra scaling issues, but I'm not. I'm pointing out that everyone has them especially pre-AI, and all of those reasons are on the table, not less likely. You have done the opposite: committed to a suspicion. That is the end result of the thought-terminating cliche.
[0] https://github.com/resources/insights/ai-powered-workforce-p...
Or it could be that GitHub saw a 14x increase in commit volume last year[0], and we've concomitantly seen its reliability fall of a cliff in the last year or so. Given that Microsoft is leasing additional space on AWS(!)[1] to handle the additional commit volume, my personal money is on commit volume growth being a bigger issue than internal use of AI.
Internal use of AI may have been an issue. Commit volume growth may have been an issue. Unless one has direct knowledge of their infrastructure issues, claiming to know is quite literally making exactly the "they are vegan, their illness must be caused by their veganism" argument the GP commenter was talking about.
[0]https://daringfireball.net/linked/2026/05/04/commits-on-gith...
[1]https://www.businessinsider.com/microsoft-github-amazon-ai-c...
"Vegans and vegetarians may have higher stroke risk" - https://www.bbc.com/news/health-49579820
"Vegans had a 43% higher risk of fractures overall compared to nonvegetarians, as well as higher risks of hip, leg, and vertebral fractures." - https://sniglobal.org/plant-based-diets-and-fracture-risk/
"The Impact of a Vegan Diet on Many Aspects of Health: The Overlooked Side of Veganism" - https://www.cureus.com/articles/138315-the-impact-of-a-vegan...
"..people who followed a vegan diet had noticeably low levels of iodine in their bodies, an element that is essential for growth, bones, and brain function. In addition, vegans had lower bone health scores..." - https://www.bfr.bund.de/en/press-release/vegan-vegetarian-be...
So indeed, the "it must be veganism" is not an unfounded concern when health complications arise, in a very similar way to "it must be the AI" is a valid concern when software issues arise.
But I was more getting at, say, staying out of the sun or being skinnyfat as a vegan, and suddenly you look "sickly"/"frail" when you'd be given the grace of looking like most people otherwise.
A similar analogy would be someone saying "well, of course you do" if you have any malady while having been vaccinated. My point being to bring up the thought terminating cliche of it compared to doing the necessary further analysis to link the malady with the suspected cause.
---
> "Vegans and vegetarians may have higher stroke risk"
It was a lump vegetarian + vegan group with a weak CI bounded at 1.02 for 3/1000 cases over a decade. The same group also had a more robust benefit of less heart disease than meat eaters. The stroke outcomes aren't replicated in other cohorts either, afaik. But the heart disease benefits are.
> "Vegans had a 43% higher risk of fractures overall compared to nonvegetarians, as well as higher risks of hip, leg, and vertebral fractures."
The study used a single baseline questionnaire for 17+ years and looked at vegans with correctable nutrition deficiencies to see +15/1000 hip fractures over 10 years. I'll grant that a poorly planned diet, especially 30 years ago with less nutritional understanding, has worse health outcomes. Just like I wouldn't use the average American's diet to lambast an omnivore diet (compared to, say, the "Mediterranean" diet).
> "vegans had lower iodine, bone health scores" (RBVD study)
On bones: p=0.02 in 72 people with 5% less QUS score in their heel bone (not DXA nor bone density tested). No body weight mediation nor data about health outcomes like fractures, osteoporosis, and no time dimension since it was just a snapshot (cross-sectional).
On iodine: It's a surrogate biomarker from a single pee test. Study didn't look at iodine-related health outcomes like thyroid dysfunction, goiter, or clinical consequences.
---
Layer-wise, the app is pretty far removed from request routing to GPU pools.
Instead, the whole system just shits the bed, catastrophically.
This is true when you have like one failure a year, but Anthropic is starting to look a lot like github lately when it comes to uptime.
After a certain point the reputation for unreliability starts sticking to you, especially when you position yourself as an indispensable tool for completing work people need done.
I'm not saying Anthropic isn't to blame for a system that is literally approaching one-nine uptime; they certainly are. I am saying that jumping to the "it must be vibe coding's fault" is an emotional confirmation-bias belief, not an evidence-based belief.
You mean like returning 529s and operating with reduced QoS?
But that's really not what they have. They have AI experts who are creating incredible LLMs.
Everything else is more than meh: Claude Code is really bad. Such a turd would never have gained any traction if it wasn't for the LLMs behind it.
I use LLMs to code daily (Claude Code still, mind you, for I didn't take the time to switch yet) and these modesl are both amazing and pathetic.
If you don't verify everything they output, they do the absolute craziest thing imaginable.
One example is I got an Anthropic model notice a "pattern" in range bound integer values. I had them range bound between, e.g., 0xCAFE0000 and 0xCAFEFFFF. And at some point a comparison/validation was needed and instead of doing an integer comparison the Anthropic model went ballistic: instead of doing an integer comparison it converted the numbers to a string, then started doing substring matching on "0xCAFE" and went even more "expert" by verifying at which position the match was happening. All that while explaining why it couldn't possibly fail.
Why did it do that? Very likely because, in a comment, it saw "0xCAFE..." as a string. And the thing saw a pattern.
Can you believe it? There's a pattern. So it must light up connections. We've got a pattern!
Now amount of kludge, hidden pre-processing, hidden post-processing is fixing the "quality" of the code produced by something that, instead of doing an integer comparison, converts things to string and then does substring searches and indexes computation.
There's no fixing that.
Yesterday: had to use three guard clauses before pushing data... Two of the three "logic gates" (as the model would explain they were, which is kinda right) he got right. The third one: same thing... It was planning to go ballistic, introduce countless lines of code, insane abstractions, to make a test that was solved with a one line timestamp comparison.
It's because it does things like that that the people who explain that they don't code anymore are delusional if they think this gives, as of today, quality code.
It's like that other dude who was happy to produce 37 K LOC per day and counting.
> ... it really says something about the quality of the world's best agentically produced code
Oh it is totally shit code. But if you monitor everything and vet everything they do, it's helpful.
I find these LLMs way more helpful at finding the source of bugs (not fixing them: finding them, which is 90% of the job anyway) and at acting like rubber-ducks then at writing code.
Claude Code sucks. Claude Code CLI sucks. Their only "solutions" to all problems is to create VMs, headless browsers, and resort to incredible hacks (the infamous "game loop" that modifies the characters output by the LLM is just shameful) etc. to try to hide the misery. It's miserable kludges everywhere.
And the only reason these miserable kludges are not entirely falling apart is because they rest on the shoulders of actual giants: projects like Linux, QEMU, etc. that were not vibe-coded.
It's sad to have useful tools (the models) and to make such poor use of them.
I'm pretty sure that, in the end, it's just like open-source powering the entire world by now: we'll have open-source projects like Pi and then newer ones that are going to come out and fix the mess we have now. And they're not going to be 100% vibe-coded by people whose jobs is "to write loops".
Goes to show how fake this industry has become when VC dollars have flooded it.
Somehow it is fine to vibe code infrastructure or security because someone (with a clear vested interest) wants you to spend more tokens at their casino because that is how they "win" at the casino (which they work at).
Except in reality, this part of software is critical and irresponsible to 'write loops" and we all know that he doesn't believe what he is saying.
Don't use it. Maybe wait a few more years. If it's not valuable/useful, then not using it, while everything matures, will not be a problem.
What can your company do?
Hire some Developers?
Are you no one?
I'm not using AI coding tools yet, and even if they force me at gunpoint to use them at work no one can force me to in my spare time
I'm not too worried about the case where no one can code anymore because that will be after I'm dead
Thirty years ago, you had an OS and you installed applications. No problem.
Later, you had to build and use apps on the internet, an infrastructure that is susceptible to DDOS attacks, government firewalls, and other security risks. Still fine, sort of.
Now, you not only have to build apps on the internet, you also have use LLMs to build apps to remain competitive with other developers. Future (human) maintainers of your code might not properly understand how it works, and if the providers of the LLMs screw up or go rogue, you are properly fucked.
There is a dependency/technology stack debt that is creating risks that need to be acknowledged.
And I know like one guy who does use them. He's not a developer by trade, he just has to write programs sometimes.
Use an Anthropic competitor?
what does someone do when a certain brand coffee maker keeps breaking; they buy a different brand.
Hard to do when each individual provider wants to lock your company into multiyear enterprise contracts.
400k isn’t crazy for the FANG set but it’s still a subset of the developer market and hundreds of thousands of those jobs have been cut in the last few years as they all collectively work to lower SWE pay.
60k a year it needs to be a full irreplaceable part of the infrastructure for I think. There are very few kinds of software that meet that bar right now (certain design tools etc that have no replacement). 12k/year is in the expensive but reasonable for the right tooling category (Matlab etc.).
I don’t know what the future holds. I know the big AI companies are banking on being able to charge for a replacement SWE that works 24/7. Still not convinced these are it yet, as useful as they can be under the right circumstances.
There aren't many tools that remain useful at that rate.
We are back to the baseline. The availability of our tools isn't adding anything in the long term because the productivity increase we get from the tooling is negated by the time we're back to doing it the old fashioned way due to downtime, so there is no claimed productivity increase espoused by the pontificators of the tooling.
uh
https://news.harvard.edu/gazette/story/2025/09/whats-driving...
https://literacybuffalo.org/2025/01/23/adult-literacy-rates-...
https://ourworldindata.org/grapher/cross-country-literacy-ra...
Anthropic has massive capability issues due to massive user growth. It happens often when EU and US work hours collide. They have smart people working on it. Don’t waste your energy complaining.
Cheers
Yeah AI Data Centers do that....
[1] - "...Data centers are projected to surge from consuming 4% of U.S. power to 9%–12% by 2030. Morgan Stanley forecasts that the resulting emissions from this American infrastructure spike will equal roughly 40% of the entire U.S..."
[1] - https://opportunitygreen.org/data-centres/?tab=overview
ps. if you say you still capable of developing software without the Internet, you're lying. Perhaps, to your own self.
I wouldn't like it and it'd be slower, but I still understand my environment in sufficient depth to work without external info if I absolutely have to. Even with AI, once in a while I ask it to just give me some hints instead of solving something for me, so I'm forced to do the work.
Me too, but let's be honest, I'm not talking about "Hello world!" experiments, I'm talking about developing usable software. I'm pretty sure, you won't be patching a Linux kernel driver on your own machine without googling stuff.
I've learned to code years before the Internet, but we've had it for so long, I'm honestly not sure anymore if I'm truly capable of building [real] stuff while offline. And I can't just ignore it, there's a feeling now, that with AI advancements, I may soon no longer be able to code efficiently without any AI.
Claude is down....
:)