I'm not sure this is even conceivable, but... Imagine the bootstrap problem of intelligence systems going down, when no one knows how to bring them back up without the system itself walking them through it... Like if power and intelligence were to go down in a Carrington-class event in 10 years...
Like what would happen to individuals ten years from now who are trying to live if AI suddenly became inaccessible
Of course we all know most humans are useless and unable to survive outside our culture, but the difference is that I'm trying to picture is the individual trying the reason through challenges. The version of this that has been obvious to me up to now is that the individual didn't have the knowledge to look after themselves. But I guess it feels different to conceive of them not having the ability to fully reason on their own...?
To not just lack the objects of intelligence, but the very process of it...
The general drove on. "At the present time our chief bottleneck is the fact that missiles are limited in intelligence. The computer controlling them can only be so large, and for that reason they can meet the changing nature of anti-missile defenses in an unsatisfactory way. Few missiles, if any, accomplish their goal, and missile warfare is coming to a dead end, for the enemy, fortunately, as well as for ourselves.
"On the other hand, a missile with a man or two within, controlling flight by graphitics, would be lighter, more mobile, more intelligent. It would give us a lead that might well mean the margin of victory. Besides which, gentlemen, the exigencies of war compel us to remember one thing. A man is much more dispensable than a computer. Manned missiles could be launched in numbers and under circumstances that no good general would care to undertake as far as computer-directed missiles are concerned . . ."
We don’t even need to imagine that world, Russia is doing it today.
Missiles and ammo are too expensive, so they send people to run at the enemy, revealing their firing positions. Then they drone the people who revealed themselves.
If I was an AI, I would reason that a T-800 is far more valuable than a man soldier who I must feed and clothe and heat. Use the disposable soldiers first and save the robots for later.
sloppy (that is, negligent) writing. Something is either (in)dispensable, or is not. Either use something that can take degrees of comparison, like valuable (A man is less valuable etc) or reword it: A computer is indispensable, a man is not.
No - the computer is not indispensable. The computer is just harder to replace than the man (in this imagined situation).
Hence - the man is much more dispensable...
---
The water in my water bottle is much more dispensable than the toothpaste in my toothpaste tube. But both can be dispensed (in both meanings - distributed or replaced).
> “There is no such adjective,” Scalia said. “I know we have used it, but there is no such adjective as choate. There is inchoate, but the opposite of inchoate is not choate.”
I'm not sure what this quote is from, but I almost down voted you in disgust before realizing that was exactly the intended effect. In other words, "Thanks, i hate it"
Sounds like a bad novel. In the year 2050 no one knows how to actually program anymore. Then ClaudeGPT by GoogleSoft goes down. A 74 year old former software developer is on his farm tending to hens.
A self driving car crashes into his barn.
Out comes out a tech bro, “Yo OG programmer bro, it’s all busted bro, could you like make it unbusted bro.”
Then again with social security set to go bankrupt it’s not like we’ll be able to retire
> Then again with social security set to go bankrupt it’s not like we’ll be able to retire
I don't know about you, but my notion that my social security check is all I will need to survive my retirement years died when my father started drafting his own social security in the 2010s, and it was literal peanuts after being a small business owner his entire life (restaurants) and paying more than his fair share in taxes. Do the average folk still have this notion that social security = retirement? Hopefully not, it's certainly misguided now, but it was misguided then, too.
I think the reason that intelligence is so scary is how contextual, stateful, path-dependent, and coupled it is.
Which doesn't seem like it. The ideal is to keep removing couplings. But in practice, I find that focusing on the statefulness, the distribution-specificness, these "accidentals", is how you actually get leverage.
The obvious social consequence is now the tendency towards centralization, tyranny, "you people can't think for yourselves, not because you have some kind of 'inherent intelligence defect', but because the world is so complicated and interconnected that only coherent systems thinking (conveniently administered from one central point) is what can save you". In the Enlightenment world, the ideal was that every man could in principle, think for themselves.
A good ending might be that everybody benefits each other by distributing intelligence, but in practice centralization always seems to win.
yeah this was the devops promise too. Enterprise evolved into dedicated teams to manage AWS and other infra, instead of reducing headcount it raised it. so it goes.
> Enterprise evolved into dedicated teams to manage AWS and other infra, instead of reducing headcount it raised it. so it goes.
I wager that's mostly due to the Jevons paradox, though; it's so much cheaper and easier to build out systems with cloud than on-prem that you end up building way more systems than you would have without it, resulting in an overall headcount increase.
My fear was that people like Linus would die and then nobody would be able to build operating systems anymore, because zoomers and below can’t even use computers.
Now we know they won’t even be able to read and write. But maybe Claude sonnet 12817 can maintain the operating system. It’ll be 17,000 dependencies in and mostly written in JavaScript, but it’ll work.
Hardly a unique problem to AI systems. You certainly can't bootstrap pretty much any industry to its current state without the existing items of its current state. Any sort of chip fab is a great example. Or most manufacturing. Or even more fundamental: mining (both metals like iron, and energy like coal or gas).
You go way too far - the leap you have made from some sort of AGI to "now individuals don't have the ability to fully reason on their own" is not clear to me at all.
If people are spending less time bent over monitors or at desks under the dictates of corporate imperative, they will probably be spending more time doing actually interesting things. Reasoning is innately human. What we call "knowledge work" is not and is often anything but. People who are naturally curious will still discover things, even if irrelevant (from a market perspective) when the AI is on, for their own satisfaction.
If such an event were to occur, they would be called upon. It's really little different from what would happen today if suddenly the power grid or supply lines were down for a sustained period of time.
The hordes of acolytes mumbling “Jevon’s paradox, Jevon’s paradox” every time AI gets cheaper beg to differ. Every minute saved using AI is just an opportunity to use more AI
People will have more things to worry about sooner. Where I live, if the power goes out so does my water. If AI is involved in either at any time in the future we will deserve what happens to us.
Also I disagree with the permise. People learned how to make technology that never existed once already, they can learn to do it again if needed. It will take time, which we'll have plenty of because your phone won't be working.
It's a real problem but your framing of it is itself very alarming.
Most humans aren't useless by any measure of most or usefulness. Our culture is our tools, dying due to a lack of antibiotics doesn't mean you didn't have anything to contribute, it means you didn't have access to the tool you needed. This is the problem we need to guard against.
No one is going to lose the ability to reason! This is accepting the most extreme version of the marketing presented by frontier labs, don't do it. People may lose fluency or familiarity or access to training on specific skills. We should put some care into making sure that can't be a catastrophe, for sure.
You've obviously never stood in a room of college freshmen and tried to deal with a young man having repeated issues being able to type his _own_ name into a box on a screen.
This is basically the premise of Bacigalupi's "Pump Six", from "Pump Six and Other Stories".
The titular Pump Six breaks down and basically there's nobody left in the world who knows how to even reason about the steps needed to fix it. There's just some log dump and nobody understands what it means. Humans are no longer capable of complex reasoning, they just live in the moment and hump.
> Of course we all know most humans are useless and unable to survive outside our culture, but the difference is that I'm trying to picture is the individual trying the reason through challenges. The version of this that has been obvious to me up to now is that the individual didn't have the knowledge to look after themselves. But I guess it feels different to conceive of them not having the ability to fully reason on their own...?
What this misanthropic/Anthropic-boosting bait is missing is that all of these pieces falling together doesn’t make sense.
1. We are so accustomed to not-even-thinking means that we are legitimately freed from labor and cared for. (The dystopic version of being freed from your labor means that you still have to constantly think to survive. But not in any “economically useful” ways. Surviving for yourself and your family outside of the robot economy)
2. At the same time as point 1, we voluntarily, with all of our leisure time, just don’t think at all. Where is the self-actualization? Where is the spontainety? The drive? Did this in all of these years not come back after being a rat racer/wage slave?
Curiously, the “hacker ethos” here is to create. Not to just “consume”. And that is taken as an intrinsic thing, not just a way to get rich. But that’s not how “people” are?
Other people are mostly like you. But maybe those here who spent their careers implementing Entertaining Ourselves To Death maws genuinely think that “people” (not them though) would genuinely, when left to their own devices and having beeng catered to every need, would just “consume” mindlessly onto death.
Sorry to Goomba anyone here.
> Of course we all know most humans are useless and unable to survive outside our culture,
All. Take away the “habitat” from an animal and they are helpless, useless, impotent.
We hope an average guy who was stuck in an experimental cryopod gets accidentally woken up and through a course of events gets appointed to the presidential cabinet long enough to fix society.
There is a TV show called "Alone" where people have to live in a remote area using basic survival skills and very minimal tools. It's interesting, but I've always wanted to see a version of that show where small communities are competing against each other.
If we practiced for these Armageddon scenarios what would a highly prepared group of 20 be able to accomplish in a year. How fast could they get humanity back to the bronze age?
Imagine having a bunch of Claude agents doing time sensitive work and this happens and now a single human needs to pick up several 8k PRs authored by bots that can't talk to you now. Not sure how effective it would be to switch models using OpenRouter and the like given that, as the Earendil guys said, there is a lot of session detail that the provider doesn't share. The uptime is really going to cap their market penetration imo.
> Imagine having a bunch of Claude agents doing time sensitive work and this happens and now a single human needs to pick up several 8k PRs authored by bots that can't talk to you now.
8k diffs PR is incredible, is it even sane for a human to comprehend that much?
There has to be a name for such scenario were the handoff from Agent to Developer is so difficult, its not even worth pursuing but to instead wait until the Agent comes back online.
No matter how good of a developer you are, we are still constrained to our minds working memory. Lots of people I have spoken to are aware of this. Agents produce lots of diffs in a fast pace, our minds don't get the time to grasp the changed logic. And then, we an Agent is unavailable all of the sudden, say in the case of it being an outage or usage limit reached, the human now have to go through massive amounts of changes. They cannot continue as usual, there is a spike, they have to learn the new changes and grasp everything before they can continue towards the goal.
If you're not using Codex and Claude at the same time, you're missing out heaps. Even without outages, OpenAI models are great reviewers for Claude; and vice versa.
Other than trivial PRs; everything I do with Opus/Fable gets reviewed by Astra; and everything I do with Astra gets reviewed by Opus/Fable.
Using only models from a single vendor, is like testing your website/webapp only on Chrome.
For me, commits are the atomic block; and a good commit should be self-contained anyway; where no additional context is necessary. If an agent can't figure out what a commit is supposed to do, with only the commit title/description and diff, it's a bad commit, and this has always been true in software engineering. I do not pass prompts between the two ordinarily.
After claude or codex finishes a commit, I switch console tabs and ask the other to review it; with the commit ID. I often find it helpful to inject a bit of human knowledge, and callout any areas of attention I see from a quick skim. (But that could just be me wanting to not abstract myself away from software engineering that much :)
Sometimes, I do ask Codex to read my ~/.claude/; and vice-versa. But, generally, I try to keep as much knowledge (e.g. investigations, reports, deep dives) inside the git tree as possible; so that is not necessary.
I don't use skills, but I do have ~/.codex/AGENTS.md and ~/.claude/CLAUDE.md. These are high-level instructions for what (1) I consider readable, maintainable code and patterns, and (2) workarounds for empirically observed model behavior IOdon't like, such as Astra being a bit of a "over-correct over-validation nit-picker". I keep these human-authored, and update regularly based on what I find annoying.
I do all of this before I submit a PR; but of course, for trivial stuff (e.g. CSS changes, copy/string changes, etc), I don't bother.
My practices and workflows do change over time. Back in the ~Opus 4.5 days I'd often define a rubric/criteria in a markdown file, iterate with AI to improve it, and that's the "spec". I've stopped doing that since GPT 5.6; partly because models have gotten a lot better at understanding high level intent from the context; and partly because nearly all models these days feel 'gradermaxxed' when working like that.
Finally, consumer $100/$200mo subs get you _so_ far, I get a lot of value from both. I used to have multiple Claude subs for a while, but trying to 'get full value' made me work on projects just for the sake of it; so 2x$200/mo is my cap :)
There's far more than zero evidence, see https://arxiv.org/html/2402.08806v1 ; or the industry-standard practice of using multiple model families as LLM judges; or even 1P implementations https://code.claude.com/docs/en/advisor (which misses most of the benefit; since you want different model families, with different pretrains and posttrains).
Never thought of it, but knowing that they encrypt reasoning is it possible that they share the decryption keys with other cloud providers like GCP Vertex / AWS Bedrock?
Neither electricity, nor the Internet is provided by one or two organizations globally, both with a blast radius that quite often takes out all of their services.
Not sure how long the latter will remain true – centralization is well underway there; try not to think of Cloudflare too much...
Globally no, but in the US there probably at most 2 providers covering your area + 5G/LTE providers. Electricity - single provider in the area. So if your job is on-site, then entire company stops working when either goes out. You don't really care that another company on the other side of the globe or street continues working.
We already got used to "internal going down globally" when AWS was the main cloud provider. Now its 3, AWS, GCP, Cloudflare, but "portion of internet that you care about is down for the entire world" is a casual thing to happen now.
When I worked at one of the early paging and cellular companies, every company had phones and pagers on at least one or two competitors for the oncall folks - you can't get a call to fix your cellular network when you are down.
Likewise while at Google, they didn't let us put Google Voice numbers down as our oncall phone number. Some folks even had multiple phones on different networks as backups.
I'm sure they have either other SOTA models or open source models around to help if things to completely pear shaped.
You can get away with a lack of rigor in software engineering (from both humans and AI agaents) because it's often cheap to try it and see. Once you get to operating the platform, it changes, and operational decisions need to be made carefully and code needs to be deployed carefully.
Oh no, you can pretty obviously just keep pushing out slop that only half works. Your customers will grumble and eat it like the peons your C suite thinks they are.
For whatever reasons, customers just buy trash nowadays, put up with trash, ignore horrifying contract terms, buy more trash, grumble about the trash, and then invest more in the trash.
I've used some engineering programs that were appalingly buggy for any application, never mind how much they cost - but once you managed to fight through the bugs, they did the thing they were supposed to reliably, and you could trust the results.
Unfortunately, not a mystery; it's management. As a syseng I can build you a 100% reliable system. But it'll fall to pieces when "Management" takes away our power to require change control, quality control, guardrails, rollbacks, performance testing, system limits, least privilege, etc. Any time you see a big company go down, it's because nobody was willing to enforce what's needed for reliability.
Given the amount of usage I'm now getting from Opus 5.5, I'm surprised it took this long since release for their first outage. In my experience, it's about 10x the usage of Fable 5.1 with similar, perhaps greater levels of intelligence.
Anthropic yesterday: We have perfect AI agents to do everything now, value us at $100 Trillion
Anthropic today: We don't really have AI agents, they are glorified versions of automation scripts and one has broken today. But tomorrow we will have perfect AI agents, so still value us at $100 Trillion.
A more cynical version of me could imagine scheduling an outage to make people think about how dependent they are on your product in advance of the IPO
Yesterday got email from Claude that they are offering 50% off for one month if I renew my subscription. I bought it today and now this. I wonder if they sent that email to too many people.
Rumours have it that the migratory flutterings of end users from Anthropic to OpenAI (sometimes taking a wrong turn and entering Grok purgatory), then back to Anthropic a season (2 weeks) later, so on, moves the great Cords of The Sky that crank the sun and moon into place, which turns the great wheel of the seasons.
If you listen closely, you may hear them squawking in annoyance after each announcement.
I've seen many people with 500s for over an hour, meanwhile the status page was green with no issues whatsoever.
Have they tried using a fraction of their immense budget to run E2E tests?
At this point downdetector is more reliable than any of these "status" pages. And it does show reports starting 9:30ish AM EST.
I've been getting 503s from the API. When I first checked everything was green, then it said "Claude.ai (including the desktop and mobile apps), Claude Code and Claude Cowork" and I see now they've included the API in the failure as well
Project completely disappeared. Conveniently for Anthropic, the bug report also went down, so they weren't spammed by my stale complaints after the project magically resurfaced.
Maybe an increase in sign ups due to people learning of their existence from the IPO? It's the authentication that seems to be failing, not the actual inference.
Elevated errors on claude.ai, Claude Code, Claude Cowork and the Claude API
Subscribe
Update - We are investigating reports of degraded performance affecting claude.ai, platform.claude.com, the Claude API, Claude Code, and Claude Cowork. We will provide an update as soon as possible.
Sep 29, 2026 - 14:30 UTC
Update - We are investigating elevated error rates affecting Claude.ai (including the desktop and mobile apps), Claude Code and Claude Cowork. Some requests to the Claude API are also returning errors. We will provide an update as soon as possible.
Sep 29, 2026 - 14:28 UTC
Investigating - We are investigating elevated error rates affecting Claude.ai (including the desktop and mobile apps), Claude Code and Claude Cowork. Users may see failed requests, errors loading or sending conversations, or be asked to sign in again; retrying may succeed. We will provide an update as soon as possible.
Sep 29, 2026 - 14:21 UTC
"Update
We are investigating reports of degraded performance ... We will provide an update as soon as possible.
Posted 1 minute ago. Sep 29, 2026 - 14:30 UTC
Update
We are investigating elevated error rates ... We will provide an update as soon as possible.
Posted 2 minutes ago. Sep 29, 2026 - 14:28 UTC
Investigating
We are investigating elevated error rates affecting Claude.ai ... We will provide an update as soon as possible."
just a hunch, but i think they might just provide an update as soon as possible.
Berget's "Cached input tokens are billed at the same rate as regular input tokens" pricing model is kind of painful for lots of coding use-cases, especially when competitors are offering significant discounts for cached input tokens.
I'm not sure this is even conceivable, but... Imagine the bootstrap problem of intelligence systems going down, when no one knows how to bring them back up without the system itself walking them through it... Like if power and intelligence were to go down in a Carrington-class event in 10 years...
Like what would happen to individuals ten years from now who are trying to live if AI suddenly became inaccessible
Of course we all know most humans are useless and unable to survive outside our culture, but the difference is that I'm trying to picture is the individual trying the reason through challenges. The version of this that has been obvious to me up to now is that the individual didn't have the knowledge to look after themselves. But I guess it feels different to conceive of them not having the ability to fully reason on their own...?
To not just lack the objects of intelligence, but the very process of it...
The general drove on. "At the present time our chief bottleneck is the fact that missiles are limited in intelligence. The computer controlling them can only be so large, and for that reason they can meet the changing nature of anti-missile defenses in an unsatisfactory way. Few missiles, if any, accomplish their goal, and missile warfare is coming to a dead end, for the enemy, fortunately, as well as for ourselves.
"On the other hand, a missile with a man or two within, controlling flight by graphitics, would be lighter, more mobile, more intelligent. It would give us a lead that might well mean the margin of victory. Besides which, gentlemen, the exigencies of war compel us to remember one thing. A man is much more dispensable than a computer. Manned missiles could be launched in numbers and under circumstances that no good general would care to undertake as far as computer-directed missiles are concerned . . ."
We don’t even need to imagine that world, Russia is doing it today.
Missiles and ammo are too expensive, so they send people to run at the enemy, revealing their firing positions. Then they drone the people who revealed themselves.
If I was an AI, I would reason that a T-800 is far more valuable than a man soldier who I must feed and clothe and heat. Use the disposable soldiers first and save the robots for later.
>A man is much more dispensable than a computer
sloppy (that is, negligent) writing. Something is either (in)dispensable, or is not. Either use something that can take degrees of comparison, like valuable (A man is less valuable etc) or reword it: A computer is indispensable, a man is not.
No - the computer is not indispensable. The computer is just harder to replace than the man (in this imagined situation).
Hence - the man is much more dispensable...
---
The water in my water bottle is much more dispensable than the toothpaste in my toothpaste tube. But both can be dispensed (in both meanings - distributed or replaced).
Wow you and Justice Scalia should have met!
> “There is no such adjective,” Scalia said. “I know we have used it, but there is no such adjective as choate. There is inchoate, but the opposite of inchoate is not choate.”
https://www.abajournal.com/news/article/justice_scalia_deliv...
Take it up with Isaac Asimov, I guess.
I was unfamiliar with the short story “The feeling of power” but thought for sure that “graphitics” must’ve been an Asimovism
I'm not sure what this quote is from, but I almost down voted you in disgust before realizing that was exactly the intended effect. In other words, "Thanks, i hate it"
asimov, "the feeling of power"
Thank you!
Asimov rocks! You should look into "Profession" and "It's Such a Beautiful Day", if you haven't.
Some of us are here to help. $1000/hr.
Joke's on you, after the AI economic collapse in a few years $1000 will have the purchasing power of 50 cents of today's money.
One meeeelion dollars then.
You've just given them a 50% discount
Sounds like a bad novel. In the year 2050 no one knows how to actually program anymore. Then ClaudeGPT by GoogleSoft goes down. A 74 year old former software developer is on his farm tending to hens.
A self driving car crashes into his barn.
Out comes out a tech bro, “Yo OG programmer bro, it’s all busted bro, could you like make it unbusted bro.”
Then again with social security set to go bankrupt it’s not like we’ll be able to retire
Thinking more a mashup of The Matrix and Idiocracy with Doc Brown as the Hero retired Programmer. Or Rick.
Hero ?
What if the twist is that he’s writing spaghetti code that only he knows how to maintain.
Worse, it’s in a language that his llc charges licensing fees to use!
If they turn it into a movie, they would send the greybeard to space to fix the AI satellite server in space to save humanity.
This is the type of movie I would be so excited about until I found out it was made by Seth Rogan.
How about Adam McKay, then?
> Then again with social security set to go bankrupt it’s not like we’ll be able to retire
I don't know about you, but my notion that my social security check is all I will need to survive my retirement years died when my father started drafting his own social security in the 2010s, and it was literal peanuts after being a small business owner his entire life (restaurants) and paying more than his fair share in taxes. Do the average folk still have this notion that social security = retirement? Hopefully not, it's certainly misguided now, but it was misguided then, too.
Social Security is more than enough in Vietnam or another low cost of living country.
The issue then is getting a visa.
Social security has a low maximum, if your dad was used to clearing 15k a month and then social security only paid him 3500$, that’s rough.
For the typical person only making around 4500$ , it’s an adjustment, but doable
A few of us old boys can still maintain Cobol
Right behind you with fortran.
A safer currency and rate would be $1MWh/hr
Talk about a #MeToo movement :)
Exactly what I was thinking.
I think the reason that intelligence is so scary is how contextual, stateful, path-dependent, and coupled it is.
Which doesn't seem like it. The ideal is to keep removing couplings. But in practice, I find that focusing on the statefulness, the distribution-specificness, these "accidentals", is how you actually get leverage.
The obvious social consequence is now the tendency towards centralization, tyranny, "you people can't think for yourselves, not because you have some kind of 'inherent intelligence defect', but because the world is so complicated and interconnected that only coherent systems thinking (conveniently administered from one central point) is what can save you". In the Enlightenment world, the ideal was that every man could in principle, think for themselves.
A good ending might be that everybody benefits each other by distributing intelligence, but in practice centralization always seems to win.
This is why it will be more important than ever to have experienced, quality employees.
I dare companies to gamble on running their businesses on absolute skeleton crews long term
yeah this was the devops promise too. Enterprise evolved into dedicated teams to manage AWS and other infra, instead of reducing headcount it raised it. so it goes.
> Enterprise evolved into dedicated teams to manage AWS and other infra, instead of reducing headcount it raised it. so it goes.
I wager that's mostly due to the Jevons paradox, though; it's so much cheaper and easier to build out systems with cloud than on-prem that you end up building way more systems than you would have without it, resulting in an overall headcount increase.
Keep an offline copy of Qwen or something together with a knowledge base. Easy if you plan ahead.
Yeah seems unlikely we'll be without at least a baseline level of AI unless we entirely run out of energy and components
My fear was that people like Linus would die and then nobody would be able to build operating systems anymore, because zoomers and below can’t even use computers.
Now we know they won’t even be able to read and write. But maybe Claude sonnet 12817 can maintain the operating system. It’ll be 17,000 dependencies in and mostly written in JavaScript, but it’ll work.
Who needs Sonnet 12817 when 5 is maintaining is already maintaining Omarchy? And it's only 5% JS!
Hardly a unique problem to AI systems. You certainly can't bootstrap pretty much any industry to its current state without the existing items of its current state. Any sort of chip fab is a great example. Or most manufacturing. Or even more fundamental: mining (both metals like iron, and energy like coal or gas).
You go way too far - the leap you have made from some sort of AGI to "now individuals don't have the ability to fully reason on their own" is not clear to me at all.
If people are spending less time bent over monitors or at desks under the dictates of corporate imperative, they will probably be spending more time doing actually interesting things. Reasoning is innately human. What we call "knowledge work" is not and is often anything but. People who are naturally curious will still discover things, even if irrelevant (from a market perspective) when the AI is on, for their own satisfaction.
If such an event were to occur, they would be called upon. It's really little different from what would happen today if suddenly the power grid or supply lines were down for a sustained period of time.
It may sadden you to find that the majority of the population can go days, even weeks without ever seriously "reasoning" about anything major.
The hordes of acolytes mumbling “Jevon’s paradox, Jevon’s paradox” every time AI gets cheaper beg to differ. Every minute saved using AI is just an opportunity to use more AI
> live if AI suddenly became inaccessible
some weeks ago, we changed the billed credit card - something went wrong on their end we were unable to use it for ~8-9 days.
Some pepople during this period said to certain tasks: "I will care about that if Claude is available again"
So, we are already there
Like navigating to the other side of the city without Maps(Google)?
People will have more things to worry about sooner. Where I live, if the power goes out so does my water. If AI is involved in either at any time in the future we will deserve what happens to us.
Also I disagree with the permise. People learned how to make technology that never existed once already, they can learn to do it again if needed. It will take time, which we'll have plenty of because your phone won't be working.
There are many people who have the weights of open source LLMs locally and just need 120v
I really hope there will be social change - flexing mind - like flexing body in a gym/sport - becomes a thing.
There always was. Like there's a gym crowd, there's an intellectual-hobby crowd.
It's a real problem but your framing of it is itself very alarming.
Most humans aren't useless by any measure of most or usefulness. Our culture is our tools, dying due to a lack of antibiotics doesn't mean you didn't have anything to contribute, it means you didn't have access to the tool you needed. This is the problem we need to guard against.
No one is going to lose the ability to reason! This is accepting the most extreme version of the marketing presented by frontier labs, don't do it. People may lose fluency or familiarity or access to training on specific skills. We should put some care into making sure that can't be a catastrophe, for sure.
You've obviously never stood in a room of college freshmen and tried to deal with a young man having repeated issues being able to type his _own_ name into a box on a screen.
One kid is spectacularly stupid, therefore most of humanity is useless?
Your dependence on AI is my margin of opportunity...
This is basically the premise of Bacigalupi's "Pump Six", from "Pump Six and Other Stories".
The titular Pump Six breaks down and basically there's nobody left in the world who knows how to even reason about the steps needed to fix it. There's just some log dump and nobody understands what it means. Humans are no longer capable of complex reasoning, they just live in the moment and hump.
> Of course we all know most humans are useless and unable to survive outside our culture, but the difference is that I'm trying to picture is the individual trying the reason through challenges. The version of this that has been obvious to me up to now is that the individual didn't have the knowledge to look after themselves. But I guess it feels different to conceive of them not having the ability to fully reason on their own...?
What this misanthropic/Anthropic-boosting bait is missing is that all of these pieces falling together doesn’t make sense.
1. We are so accustomed to not-even-thinking means that we are legitimately freed from labor and cared for. (The dystopic version of being freed from your labor means that you still have to constantly think to survive. But not in any “economically useful” ways. Surviving for yourself and your family outside of the robot economy)
2. At the same time as point 1, we voluntarily, with all of our leisure time, just don’t think at all. Where is the self-actualization? Where is the spontainety? The drive? Did this in all of these years not come back after being a rat racer/wage slave?
Curiously, the “hacker ethos” here is to create. Not to just “consume”. And that is taken as an intrinsic thing, not just a way to get rich. But that’s not how “people” are?
Other people are mostly like you. But maybe those here who spent their careers implementing Entertaining Ourselves To Death maws genuinely think that “people” (not them though) would genuinely, when left to their own devices and having beeng catered to every need, would just “consume” mindlessly onto death.
Sorry to Goomba anyone here.
> Of course we all know most humans are useless and unable to survive outside our culture,
All. Take away the “habitat” from an animal and they are helpless, useless, impotent.
We hope an average guy who was stuck in an experimental cryopod gets accidentally woken up and through a course of events gets appointed to the presidential cabinet long enough to fix society.
There is a TV show called "Alone" where people have to live in a remote area using basic survival skills and very minimal tools. It's interesting, but I've always wanted to see a version of that show where small communities are competing against each other.
If we practiced for these Armageddon scenarios what would a highly prepared group of 20 be able to accomplish in a year. How fast could they get humanity back to the bronze age?
"The Machine Stops"
for those not aware of the reference, "The Machine Stops" was a 1909 novelette by E. M. Forster. It was very forward thinking, and so full of ideas!
On project gutenberg: first story in this anthology, https://www.gutenberg.org/ebooks/72890
PDF of just the story: https://www.cs.ucdavis.edu/~koehl/Teaching/ECS188/PDF_files/...
Imagine having a bunch of Claude agents doing time sensitive work and this happens and now a single human needs to pick up several 8k PRs authored by bots that can't talk to you now. Not sure how effective it would be to switch models using OpenRouter and the like given that, as the Earendil guys said, there is a lot of session detail that the provider doesn't share. The uptime is really going to cap their market penetration imo.
> Imagine having a bunch of Claude agents doing time sensitive work and this happens and now a single human needs to pick up several 8k PRs authored by bots that can't talk to you now.
8k diffs PR is incredible, is it even sane for a human to comprehend that much?
There has to be a name for such scenario were the handoff from Agent to Developer is so difficult, its not even worth pursuing but to instead wait until the Agent comes back online.
No matter how good of a developer you are, we are still constrained to our minds working memory. Lots of people I have spoken to are aware of this. Agents produce lots of diffs in a fast pace, our minds don't get the time to grasp the changed logic. And then, we an Agent is unavailable all of the sudden, say in the case of it being an outage or usage limit reached, the human now have to go through massive amounts of changes. They cannot continue as usual, there is a spike, they have to learn the new changes and grasp everything before they can continue towards the goal.
If you're not using Codex and Claude at the same time, you're missing out heaps. Even without outages, OpenAI models are great reviewers for Claude; and vice versa.
Other than trivial PRs; everything I do with Opus/Fable gets reviewed by Astra; and everything I do with Astra gets reviewed by Opus/Fable.
Using only models from a single vendor, is like testing your website/webapp only on Chrome.
In my workflow, I still struggle to employ both providers in a way that makes sense and doesn't require me passing data or prompts between the two.
Could you share a bit more about how you do it?
For me, commits are the atomic block; and a good commit should be self-contained anyway; where no additional context is necessary. If an agent can't figure out what a commit is supposed to do, with only the commit title/description and diff, it's a bad commit, and this has always been true in software engineering. I do not pass prompts between the two ordinarily.
After claude or codex finishes a commit, I switch console tabs and ask the other to review it; with the commit ID. I often find it helpful to inject a bit of human knowledge, and callout any areas of attention I see from a quick skim. (But that could just be me wanting to not abstract myself away from software engineering that much :)
Sometimes, I do ask Codex to read my ~/.claude/; and vice-versa. But, generally, I try to keep as much knowledge (e.g. investigations, reports, deep dives) inside the git tree as possible; so that is not necessary.
I don't use skills, but I do have ~/.codex/AGENTS.md and ~/.claude/CLAUDE.md. These are high-level instructions for what (1) I consider readable, maintainable code and patterns, and (2) workarounds for empirically observed model behavior IOdon't like, such as Astra being a bit of a "over-correct over-validation nit-picker". I keep these human-authored, and update regularly based on what I find annoying.
I do all of this before I submit a PR; but of course, for trivial stuff (e.g. CSS changes, copy/string changes, etc), I don't bother.
My practices and workflows do change over time. Back in the ~Opus 4.5 days I'd often define a rubric/criteria in a markdown file, iterate with AI to improve it, and that's the "spec". I've stopped doing that since GPT 5.6; partly because models have gotten a lot better at understanding high level intent from the context; and partly because nearly all models these days feel 'gradermaxxed' when working like that.
Finally, consumer $100/$200mo subs get you _so_ far, I get a lot of value from both. I used to have multiple Claude subs for a while, but trying to 'get full value' made me work on projects just for the sake of it; so 2x$200/mo is my cap :)
There is zero evidence for this "use multiple models" trope, IMO.
It's just feelings.
There's far more than zero evidence, see https://arxiv.org/html/2402.08806v1 ; or the industry-standard practice of using multiple model families as LLM judges; or even 1P implementations https://code.claude.com/docs/en/advisor (which misses most of the benefit; since you want different model families, with different pretrains and posttrains).
Never thought of it, but knowing that they encrypt reasoning is it possible that they share the decryption keys with other cloud providers like GCP Vertex / AWS Bedrock?
They don’t even share the reasoning keys between *models in the same model family*.
Opus isn’t even allowed to decrypt thinking blocks from Fable if you switch models mid-conversation.
https://platform.claude.com/docs/en/build-with-claude/preser...
They don't, no (last that I looked) – if you switch providers, encrypted reasoning no longer works.
The NSA wants a word.
> now a single human needs to pick up several 8k PRs authored by bots
PRs shouldn't really be reaching 8k diffs no matter who authors them.
Given that CI is turning out to be a bottleneck - and a real pain for chainedd PR's - I have my hesitations on this.
yea, a lot of "best practices" are no longer best practices
PRs shouldn't really be reaching 8k diffs no matter who authors them.
There's an ocean between what "should" happen and what actually happens in tech today.
can't you say the say with imagine "power goes out" or the "internet goes out"?
(but yes if AI is going to propagate, they need to fix their uptimes)
Neither electricity, nor the Internet is provided by one or two organizations globally, both with a blast radius that quite often takes out all of their services.
Not sure how long the latter will remain true – centralization is well underway there; try not to think of Cloudflare too much...
Globally no, but in the US there probably at most 2 providers covering your area + 5G/LTE providers. Electricity - single provider in the area. So if your job is on-site, then entire company stops working when either goes out. You don't really care that another company on the other side of the globe or street continues working.
We already got used to "internal going down globally" when AWS was the main cloud provider. Now its 3, AWS, GCP, Cloudflare, but "portion of internet that you care about is down for the entire world" is a casual thing to happen now.
No, you can't. Because power or internet doesn't go out globally.
Claude models don't go out globally either. Bedrock is fine. Azure is fine.
Solar flares would like a word.
No worries, coding is solved they said in February 2026!
Isn't that a good news!
anyone can buy a backup generator for power
They probably have an OpenAI subscription, same for OpenAI SWE, they have a Claude Code subscription
When I worked at one of the early paging and cellular companies, every company had phones and pagers on at least one or two competitors for the oncall folks - you can't get a call to fix your cellular network when you are down.
Likewise while at Google, they didn't let us put Google Voice numbers down as our oncall phone number. Some folks even had multiple phones on different networks as backups.
I'm sure they have either other SOTA models or open source models around to help if things to completely pear shaped.
They likely also have dedicated and separate clusters for their own use.
If so - for running Chinese models :trollface:
I'm sure they run every model they can get their hands on in internal clusters.
Coding may have been solved, but uptime remains a mystery.
You can get away with a lack of rigor in software engineering (from both humans and AI agaents) because it's often cheap to try it and see. Once you get to operating the platform, it changes, and operational decisions need to be made carefully and code needs to be deployed carefully.
Oh no, you can pretty obviously just keep pushing out slop that only half works. Your customers will grumble and eat it like the peons your C suite thinks they are.
For whatever reasons, customers just buy trash nowadays, put up with trash, ignore horrifying contract terms, buy more trash, grumble about the trash, and then invest more in the trash.
It will be the death of us.
I've used some engineering programs that were appalingly buggy for any application, never mind how much they cost - but once you managed to fight through the bugs, they did the thing they were supposed to reliably, and you could trust the results.
This is not true for vibecoded stuff.
May I steal this quote as if it's mine? :bow:
It's on the internet now, inside the bellies of 1000 llms.
It's everyones.
Why do you need to say it's yours to use it? :p
You are trained now, it must be yours
Excuse me but you've distilled this comment from me
Unfortunately, not a mystery; it's management. As a syseng I can build you a 100% reliable system. But it'll fall to pieces when "Management" takes away our power to require change control, quality control, guardrails, rollbacks, performance testing, system limits, least privilege, etc. Any time you see a big company go down, it's because nobody was willing to enforce what's needed for reliability.
Given the amount of usage I'm now getting from Opus 5.5, I'm surprised it took this long since release for their first outage. In my experience, it's about 10x the usage of Fable 5.1 with similar, perhaps greater levels of intelligence.
Yeah, Fable 5.1 just seems Pareto-dominated. Opus 5.5 is the first model that feels like a step change for me, personally since like 4.6
Exactly, 5.5 is the first model that I used that's better than 4.6.
Fable5.1 is Speedy
Anthropic yesterday: We have perfect AI agents to do everything now, value us at $100 Trillion
Anthropic today: We don't really have AI agents, they are glorified versions of automation scripts and one has broken today. But tomorrow we will have perfect AI agents, so still value us at $100 Trillion.
A more cynical version of me could imagine scheduling an outage to make people think about how dependent they are on your product in advance of the IPO
The good ol' Fastly approach. Have your stock pump after you go down because no one had any idea how many websites depended on you.
The auto mode classifier has been out of service more than once in last 24 hours.
Had to shift+tab into other modes and manually approve permissions like a cave man
Ah, so disabling auto-mode is their monkey rate limiter.
Yesterday got email from Claude that they are offering 50% off for one month if I renew my subscription. I bought it today and now this. I wonder if they sent that email to too many people.
Too many OpenAI $200 customers switching over to Anthropic.
Rumours have it that the migratory flutterings of end users from Anthropic to OpenAI (sometimes taking a wrong turn and entering Grok purgatory), then back to Anthropic a season (2 weeks) later, so on, moves the great Cords of The Sky that crank the sun and moon into place, which turns the great wheel of the seasons.
If you listen closely, you may hear them squawking in annoyance after each announcement.
The Samsara of AI subscriptions.
I like the cut of your jib. Do you have a blog?
A usage reset would be a much appreciated apology. :)
Hey, it's 99.9 if you look at it upside down.
6.66% uptime? Give github a few years they'll get there
Uptime of the Beast
I've seen many people with 500s for over an hour, meanwhile the status page was green with no issues whatsoever. Have they tried using a fraction of their immense budget to run E2E tests?
At this point downdetector is more reliable than any of these "status" pages. And it does show reports starting 9:30ish AM EST.
I've been getting 503s from the API. When I first checked everything was green, then it said "Claude.ai (including the desktop and mobile apps), Claude Code and Claude Cowork" and I see now they've included the API in the failure as well
Apple App Store build upload processing is also out at the same time...hmmm....I wonder if Apple is using Claude for build upload processing.
Maybe show respect to OpenAI DevDay
OpenAI has nothing better to do than to hack Anthropic?
Would be interesting to see how OpenAI and Anthropic would respond to being "accidentally" hacked by their respective Frankenstein monsters.
Wrote last year about the fact that AIs can be down, and what that might mean: https://max.engineer/failover-to-hi
I have no clue what's going on, but I currently see a total cost of $32**.**, even though I have usage credits off.
Am I looking at somebody else's usage?
If I can't use it I just sit there. That's the fallback now.
Project completely disappeared. Conveniently for Anthropic, the bug report also went down, so they weren't spammed by my stale complaints after the project magically resurfaced.
Maybe an increase in sign ups due to people learning of their existence from the IPO? It's the authentication that seems to be failing, not the actual inference.
Jup, all currently running containers got deleted on my account. Hope you guys got your stuff pushed to somewhere before the outage.
claude has outages multiple times per week, this seems to be nothing new. and this is not ok, for a premium product.
Why can they simply not add more computer?
DRAM
Seems to be working again.
Claude outages are now the new go outside and touch grass indicator
Previous: GitHub outage
>Update - All services have been operating normally since 14:59 UTC. We are continuing to monitor. Sep 29, 2026 - 15:30 UTC
I hate it when companies declare victory when there is still a fire around them.
Yeah I had a session fail and keep failing.
https://status.claude.com/ says:
Elevated errors on claude.ai, Claude Code, Claude Cowork and the Claude API Subscribe Update - We are investigating reports of degraded performance affecting claude.ai, platform.claude.com, the Claude API, Claude Code, and Claude Cowork. We will provide an update as soon as possible. Sep 29, 2026 - 14:30 UTC Update - We are investigating elevated error rates affecting Claude.ai (including the desktop and mobile apps), Claude Code and Claude Cowork. Some requests to the Claude API are also returning errors. We will provide an update as soon as possible. Sep 29, 2026 - 14:28 UTC Investigating - We are investigating elevated error rates affecting Claude.ai (including the desktop and mobile apps), Claude Code and Claude Cowork. Users may see failed requests, errors loading or sending conversations, or be asked to sign in again; retrying may succeed. We will provide an update as soon as possible. Sep 29, 2026 - 14:21 UTC
"Update We are investigating reports of degraded performance ... We will provide an update as soon as possible.
Posted 1 minute ago. Sep 29, 2026 - 14:30 UTC Update We are investigating elevated error rates ... We will provide an update as soon as possible.
Posted 2 minutes ago. Sep 29, 2026 - 14:28 UTC Investigating We are investigating elevated error rates affecting Claude.ai ... We will provide an update as soon as possible."
just a hunch, but i think they might just provide an update as soon as possible.
Someone put update-status-page.pl in a one minute cron job.
Oh come on, I just had my weekly limit reset.
Claude, deploy to prod - make no mistakes!
"You are a distinguished engineer..."
Berget Code is running normal https://berget.ai/code
It's bad etiquette to not mention you are a founder of something you are trying to plug
Berget's "Cached input tokens are billed at the same rate as regular input tokens" pricing model is kind of painful for lots of coding use-cases, especially when competitors are offering significant discounts for cached input tokens.
Who?