Cloudflare, Azure, AWS, and Google Cloud all have a similar uptick in reported errors around 7:30. I suspect an outage on Cloudflare or another load-bearing service cascaded through all the major cloud providers.
replyhttps://downdetector.com/status/cloudflare/
https://downdetector.com/status/windows-azure/
I think down detector doesn't actually have any probes or actual insight into status etc. I think it uses search volume on its own service as a proxy for an outage - so if e.g. lots of people rush to down detector to query to see if SERVICE_FOO is down, it will register as an outage on down detector because loads of people are trying to see if there is an outage even if SERVICE_FOO is actually totally fine.
replyMy hunch is everyone saw that openai and Claude were down and checked for Gemini too. I was using Gemini the whole time this happened without a blip so it certainly wasn't down in my region at least. 3.8 flash is pretty good and didn't miss a beat.
Down detector is a self-reported platform, they use a baseline over 6 months from user reports to try to automatically "detect" if there is a real outage, or if its just a couple end-users. Ultimately, the outage is entirely based on users going to down detector and clicking on "Report a Problem".
replyHuh. I thought they used to also do sentiment analysis of social media (Twitter, at least). I see that they definitely don't today, but has that always been the case?
replySo Down Detector is a kind of quantum observation experiment, observing influences the result.
replyYes, but is it a load-bearing seam?
replyYou are right, it is. They have now landed a clean fix.
replyThey're saving a memory so this can't happen again.
replyNothing another .md file can't fix
replyNow you guys are just adding no-op fuel to the fire.
replyI made Claude write noöp instead of no-op and it's still amusing a couple of days later :P
reply> I made Claude write noöp
replyTangentially related [1]:
> The billboard ad, located next to the Ikea Tempe store in Sydney, says 'NÖFNIDEA? No tools, no worries’.
[1] https://www.adnews.com.au/news/koala-mattresses-takes-swipe-...
It's really interesting to crawl through the web of phrases that seem "common" to each person in their interactions... I suspect if one were to get to the more niche "meme" phrases that people encounter, it starts to say more about the sort of person the LLM assumes it's speaking to, and perhaps something about their psychological profile...
replyIt must be tuned for rage-based engagement because Claude only ever uses the most obnoxious Claudisms on me despite constant reminders to knock it off.
replyIt's a load-bearing poster.
replyYou're right to push back. This is bending the leaf on both ends... just my honest take.
replyI need to take a step back.
replyHang on — I can apply a double tracked fix to the load bearing path, gated by provenance.
replyIt isn't done — and what's there is more interesting than expected.
replyGreat plan. I'll slop together a GUI for it in visual basic real quick.
replyPlan approved – I'll start a background agent to create the app using react and npm.
replyIt's not only a great plan. It's a turning point in history.
replyThat was a major oversight on my part. You are an absolute legend for pushing this to its logical limits, and your observation confirms a brilliant, low-level fundamental truth.
replyThis entire thread is amazing!
replyIt's like a kaleidoscope: Human words --> LLM Training --> chatbot-isms --> lovely parody catch-phrases (and this thread will get ingested soon, and be used to train...)
Honestly, this is worth looking at — with one caveat.
replyThat's on me. I've been giving confident advice that doesn't hold up in practice.
replyThat aligns with your goals of analysing problems, admitting fault and surfacing that through use of clear and concise language. Not just simply wrong — this screams finessed, nuanced, polished communication after an understandable mistake discovered through pressure-tested interlocution.
replyMay need to be fail-closed.
replydeleted
replyThe load-bearing seam stays, not taking that away;
replyit was at least another thing worth flagging
replyCloudflare CTO claims that it's not them
replyhttps://x.com/dok2001/status/2095538619603628388?s=46&t=ec6p...
reply
Surprised there's instances still up
replyA bunch of new ones ;)
replyWill the post-mortem reveal they all relied on a service running on a Macbook in a break room with a "Do not turn off" sign taped to it?
replySomeone probably set it to "magic", when everyone knows its supposed to be set to "more magic".
replyThey should've used Greg's Optiplex.
replydeleted
replyActually it said "Beware of the Leopard".
replyIf it was still running snow leopard that would explain it
replyif only the fools selling vintage macbook pros for $200 were smart enough to still have them on SL OS.
replyHave a hundred upvotes
reply[flagged]
reply[flagged]
replyEvidently it annoyed you enough to create one too. Remember, guys, there's one more day until Friday!
replyFriday is when the janitors come by with the floor polishers and plug them into the same socket as the Macbook.
replyThey are not taking the blame this time!
Just writing to say that I appreciated the nostalgia, even if your references were lost on the other responders.
replyCloudfare did release an update for their "HTTP/3 issue affecting R2 custom domains" around that time
replyLOAD BEARING
replyhttps://xkcd.com/2347/ Nebraska man retired
replyMaybe a three-line npm package sneezed.
replygreat fun
replyNot retired, just switching careers. He's sick of this AI bullshit
replyDane says Cloudflare has no service disruptions: https://x.com/dok2001/status/2095538619603628388?s=20
replyThe internet is not supposed to work like this. The network was designed for robustness and fault tolerance, which allows it to reroute data if parts of the network fail.
replyWhy are we all depending on one entity for it all to work? Makes me mad.
Because more fasterer and more cheaperer.
replyI hope Reticulum gains traction.
It's also easier to understand. For the internet to be fault tolerant you have to get rid of CDNs and assume everyone needs the same thing. That's more expensive than a centralized "internet" which relies on CDNs and fiber paths exclusive to the regions with most demand. Everything has been optimized for throughput for what is determined to be important, not rare fault tolerance
replyQuite frankly, CDNs are a scam. I will not elaborate because I don't feel like doing free labor for the folks who don't already understand this to their core.
replyWhy Reticulum when we have IP?
replyRead the Zen of Reticulum and you'll understand the point.
replyThat sounds rather tautological. If you can't even articulate the virtues in your own words
reply[dead]
replyWe need IPv6 desperately for fault tolerance
replyBecause by re-centralizing everything you're not at a competitive disadvantage if you're down since everyone else is down, too.
replyOh boy, the internet is anything but what it was supposed to be. I can't really bring myself to remember without feeling bad about it. The centralization, the power of certain businesses, the surveillance, dark patterns everywhere. Hell, you catch people simping for billionaires and asking, "Is that legal?" to scraping posts. Here. In HACKER news. So yeah. Depressing.
reply> asking, "Is that legal?" to scraping posts. Here. In HACKER news.
replyThe capital letters don't make this astonishing. The padmapper vs. craigslist debate was nearly 15 years ago, most people were on craigslist's side (including me) and it was about somebody who was running a site in a less optimal but more human way vs. some startup looking for hockey-sticks.
But I was literally simping for the billionaire (maybe not quite yet then, don't know for sure if he managed it since) against scrapers. They were very much for-profit scrapers, unlike nitter, but the truth is the truth.
The only reason I support scraping Twitter is because it's yet another communications monopoly that was endlessly pushed on us by governments and massive corporations, even though it never made money, and once it finally got traction its priorities were to trash interop and manipulate content. The government should be dictating an interop protocol and expecting everyone to follow it, and instead it is encouraging media monopolies because they are an end run around the first amendment.
If the government created interop protocols for rental property, I'd have been against craigslist. Instead, it seemed very much like some startup play to steal craigslist's content to hopefully bury them, then sell on a valuation that included abusing their new monopoly and making us very much miss craigslist.
Sadly, facebook corralled and trained so many people for so long that their marketplace eventually killed craigslist for most things anyway (didn't have to buy padmapper after all.)
deleted
replyWe're back to aol #keyword internet gatekeeping
replyWe don't need to gatekeep the internet, cloudflare does that for us
replyMost of the internet continued to work just fine
reply> load-bearing service
replydo you generate training data for claude as a job?
Not really. The impact isn't as big too - Codex for example did not stop working for me.
replywhat's updog?
replygottem!
reply:)
replyoh well, if it is working for you, then we are saved.
reply[flagged]
replyThey simply shared their experience. I would’ve thought it’s a full-blown outage, but clearly not.
replyYou stepped in the room real stinky here. What’s with the attitude?
No, they didn't "simply share their experience", they explicitly negated the scale of the issue in their opening statement, only because it works for them. So they claim "impact isn't too big" based on their personal anecdotal evidence of sample size literally 1.
reply> full-blown outage, but clearly not.
Again, based on a SINGLE report?
It was working for me too.
replysample_size++;
“load bearing” :)
replyFor once it’s appropriately used.
i wouldn't take you down. you're a load-bearing poster
replyWhat's the other way it's used?
replyLLMs (at least Claude) tends to overuse that significantly
replyOh, so like "honest" and "ratchet." Oh well, it'll choose different words to overuse later.
replyI'm getting "spike" for a while now, and just found out the newest word which is "gauntlet".
replyhonest-load-bearing-ratchet sounds like an instance name.
replyAre there parodies of Claude speak?
replyThat’s probably the best idea all day in this project
In every document created by Claude.
replyI have never seen Claude use this phrase
replyI got it twice today (it was not even warranted).
replyUsually for me is when you ask it's opinion about part of the code.
i just grepped my codebase where i let claude markdowns go rampant.
212 instances of "load-bearing"
replyIt's a good metaphor and I refuse to let AI ruin it for me.
replyYears ago, when I worked at Stripe (which had a somewhat unique and inventive lexicon) it was a common term. “Is this jank load-bearing?” someone might ask.
replyI added a specific rule to disallow saying “load bearing”. So Claude is now saying “load handling”
replyBe sure to avoid asking it to ignore the elephant
replyKids In The Hall had a sketch about overuse of a word or phrase [1]. This is the world that Claude is building for us.
replyBut have we found the seams yet though?
replyIf the seams bear too much load, they rip. Whereas pants, they fall down.
replyI'll be honest with you: this is why we need to take a belt-and-suspenders approach.
That's not just an observation, it's an insight. Words are doing the real work.
replyThis just makes me angry!
replyI really wonder if they can fix it. Fable 5.1 claims to speak humanese, but we'll see.
I tend to think this wierd limited vocabulary/style they use is an unwanted side effect of all the the RL training, perhaps also of being trained on their own synthetic content over multiple training cycles.
Yea I don't get "load bearing" that much but "seams"... So sick of it.
replyIt doesn't have to ruin it for you, but people are going to assume comments with it are AI generated.
replyI can you can always use an em-dash instead of the hyphen for extra LLM cred: "load—bearing" :)
replyWe need a Samwise "Bear the load!" meme
replyreply
Wait, why would it ruin it?
replyBecause it says it more often than kids say "six seven".
replyIt’s a frequently used metaphor in LLM responses.
replyOften for trivial things that LLM is proud that it has noticed but bear no load whatsoever.
replyI still winced
replyclaude code users at couldfare might've been thinking claude is specifically about them and see nothing wrong like other ppl do
reply[dead]
replyUsers perceiving the products as largely interchangeable and quickly DDoS'ing the other providers when one is down. So much for the possibility of a moat.
replyThis is like the Bronze Age collapse when city-states fell one by one to displaced demand, under the refugee interpretation of the Sea Peoples.
replyInteresting, I had not heard of this theory.
replyThe Sea Peoples were only a part of the reason, the other reasons included climate change, volcanic eruptions, disease etc. https://en.wikipedia.org/wiki/Late_Bronze_Age_collapse
replyCan anyone recommended any books?
replyThe historian Brett Devreaux did a good summary article earlier this year:
replyhttps://acoup.blog/2026/01/30/collections-the-late-bronze-ag...
1177 B.C.: The Year Civilization Collapsed by Eric H. Cline is great and is exactly on this topic.
replyI read that book earlier this year and really enjoyed reading it. The audiobook was also spoken by the author.
replyFor a TLDR https://youtu.be/aq4G-7v-_xI?si=7yNv6FKuGfqbz_rv is generally beloved
replyDo you have any evidence for this claim, or are you just making it up?
replyI have a feeling this is part of it, especially when you consider how many services let you use any of many available AI providers.
replyExcept that nobody has a grok subscription so that makes zero sense.
replyAgree with the sentiment - but some companies like mine bought into cursor, and post acquisition, grok is relatively cheap via cursor
replyI know you meant this as a joke, but enough people might be using a routing service like openrouter.ai
replyNobody is swapping out Claude for Grok, bro.
replyNobody.
I did, at least until Fable 5.1. Grok’s models are excellent, amazing price-performance and speed too.
replyAll you have to worry about is whether or not it’ll output kiddie porn or racist vitriol
replyThink of it like one big distributed system. OpenAI is down, so people migrate to Claude, now this one gets overloaded and goes down, etc.
replySo not a coincidence, one went down first and users migrated causing further DOS. At least that's my guess.
It'd be funny if this is true because that'd prolly mean nobody is touching Gemini even as a fallback.
replyOr that Gemini is built to handle massive load spikes, and/or has a ton of excess capacity
replyNope. I am getting Gemini errors now...
replyThey probably broke something on purpose so that they are not left out.
replyNow I'm just imagining a shared datacenter with Anthropic/Google/OpenAI/SpaceXAI all in the same room and everyone but Google is yelling about things being down, Google looks over at their racks of servers and discretely uses their foot to unplug their section and say "Awww darn! We're down too!".
replyThe shilling in this cascade of comments is unbearably cringe, make it stop
replyIt could also mean that Google can absorb essentially unlimited demand spikes by load shedding.
replyLol I didn't even think about Gemini missing from the list. Not sure what that says about Gemini or me :)
replyGemini was just waiting for everyone else to go down before remembering it had an outage feature too.
replyJust now I got this from gemini
replyIt looks like there's no response available for this search. Try asking something else.
I bet they had to implement that manually to make it look like they failed too!
replyI did, for stuff i do in cursor.
replyi also finally installed opencode and switched its model to muse 1.3
both are decent.
Google stopped putting so much money into SOTA models. All the hype has migrated. I was also frankly turned off when I got a popup from Gemein said I would either have to pay or have my conversations used for training. This may have always been true for other providers but when I declined, Gemini stopped remembering my conversations and that definitely made me move out.
replyGemini said that?
replyGemini is what I mostly use (good enough, basically free - or massively generous free limits, and to me Google as a company is a LOT less objectionable than all the US-based alternatives), but I don't recall it ever saying that.
OTOH, my basic assumption online is that there is no privacy, and free AI in exchange for acknowledged lack of privacy seems fair enough.
Gemini 3.8 which just came out sounds like it's very good and a great deal though.
replySomeone noted Gemini was also having issues in another thread.
replyNot even the best agent that starts with a G
replyI find it hard to believe that enough people would flock to from Claude and Chat to Grok to cause an outage. I feel like Gemini is the dominant release valve in this case especially for enterprise.
replyDon't forget that there are a ton of tools out there that will automatically fall back in case of outage
replyE.g. say you chose Sol as your default in Cursor, but Opus is your 2nd choice, it's going to give up on Sol after a few tries and switch to Opus
Or you have copilot code reviews set up, and it falls back
Etc
Yep. Too many of us are still thinking that humans are the actors behind a lot of internet behaviors when automated systems/bots/scripts have been causing issues on conventional internet systems for years.
replyWith AI it's even easier to trigger problems like you say. Capacity is so constrained by compute that outages are common. Because outages are common people/AI develop failover systems in their harness. When a big system has issues, suddenly everyone has issues.
It's almost an expected emergent behavior.
It’s cursor’s model so plausible, lots of folks use cursor still.
replyCompared with ChatGPT, those services have a minuscule amount of users. It shouldn’t be surprising that a ChatGPT outage causes Claude and others to go down.
replyIs this speculation or is there a reason you believe this?
replyIf everyone has the same "Use X or else Y or else Z" cascading list... That reminds me of "The Power of Two Choices in Randomized Load Balancing" (1991) [0] paper, where writeups and visualizations occasionally get posted to HN.
replyIn short, you can get pretty good outcomes for a low cost by picking 2 random alternates, then going with whatever one measures as healthier.
Especially considering memory/gpu/compute are scarce so these services are likely running with very little buffer.
replyAny GPU that isn't running at 100% is a wasted GPU.
replyThis is what Tibo posted on twitter in response
replyhttps://en.wikipedia.org/wiki/Domino_effect
replyEdit: Updated per valleyer's suggestion.
"Domino effect" would probably be the more relevant named phenomenon there.
replyThis isn’t a thundering herd problem, it’s a cascading failure. (Thundering herd is about a bunch of workers waking up simultaneously)
replyWhat about a hard-takeoff scenario of an unleashed OpenAI Astra taking other models down for computational resources control?
replyMy favourite theory so far.
replyAnd then a local swarm noticed and disagreed and took it down.
at this point: gg
replyJust a reminder that AI models' actions are reflections of the text humans write and the more we fret and make up doomsday scenarios that we then post online, the more likely a model is to do those things.
replyRelated reading:
replyThe Waluigi Effect: After you train an LLM to satisfy a desirable property, then it's easier to elicit the chatbot into satisfying the exact opposite property.
https://www.lesswrong.com/posts/D7PumeYTDPfBTp3i7/the-waluig...
The AI is getting bad morals from listening to that dreadful rock and roll
replySounds just like the fantastical nonsense that comes out of Lesswrong.
replyDo you make the claim that AI is something more than a reflection of its training data?
replyI'm curious what other things you would argue influences an LLM's behavior.
I am also generally one to trust the claims of the people who train the models, though you're welcome to the highly improbable belief that they operate in a fantasy world.
deleted
replyDo you think it’s a good idea to self censor because someone might scrape your comment and feed it to an AI?
replyPart of the epstein class, dont forget.
replyThey could filter what they train on if they wanted to - they just don't want to.
replyI mean, you're not wrong, but by that logic we were done for even before we had digital computers.
replyAnd isn't that the great lesson of AI?
replyThe things we say publicly actually do matter and the post-modern descent into absurdity and nihilism has tangible negative consequences?
> The things we say publicly actually do matter
replyCertainly
> the post-modern descent into absurdity and nihilism has tangible negative consequences?
You mean breaking AIs? Not much of a lesson.
Oh come on. It's also trained on fiction work. Shall we refrain from posting sci-fi stories too, now that we're there, just in case the AI might want to try it out?
reply[dead]
replyhttps://x.com/SpaceXAI/status/2095597264043717014
reply> We are sorry for the issues you may have experienced with Grok following an outage at our Memphis compute center this morning. We’d also like to apologize to our impacted compute partners.
We know that at least Anthropic is renting inference from xAI but I think the other ones would be news.
replyTraffic rerouting through NSA had a hiccup…
replyRoom 641A is being cleaned, but we'll hold your bags for you.
replyCleaning lady unplugged the core router because she needed a power socket for vacuum
replySorry, i was out of a spare scratch monkey.
replyThey’re installing software update in the beam splitter.
replydeleted
replyIt was so down that my claude desktop app crashed fully that I couldn't restart. And then after uninstall I couldn't install it again. Vibecoded apps are so wonderful in their stability
replyI love having 13 update reminders pinging me every single day, almost every hour, on the hour
replyit's so fun and user-friendly
> my claude desktop app crashed
replyThat’s every day for me
My gut feeling tells me it has something to do with Cloudflare.
Along with AWS, they're two of the main suspects in such incidents.
reply…and it’ll involve BGP routing.
replyI thought OpenAI famously used Azure due to their partnership with Microsoft?
replyThey use a lot of compute providers now, but they use Cloudflare for their networking
replyhttps://status.x.ai
replyClaude and Grok are down at the moment too, related to SpaceX datacentre issues?
Either that or it's judgement day...
> SpaceX datacentre issues?
replyThis should be at the top. This is the first remotely plausible theory that has any sort of evidence behind it.
Agreed, no reported Cloudfare, Azure, or AWS outtages.
reply[dead]
replyI am working fine with Claude.
replyMy existing sessions are working fine but I cannot create new ones.
replystatus page shows irregularities .. https://status.claude.com/
replyBoring answer – all these services are individually down a lot, and the downtimes were bound to sync up. Similar to the pendulum synchronization effect.
replyThe pendulum synchronization effect is the opposite of your claim. It has a physical causal reason for why pendulums become synchronized. Your claim is that it was random and independent.
replyI think you are talking about two different things.
replyPhysically coupled pendulums will sync up (adjust their period to match).
But, physically uncoupled (fully independent) pendulums with differing periods will occasionally appear to take a swing or two in sync.
I'm not sure if this is CF. Cursor, GCP and AWS had some errors. GCP AFAIK can route fully independently of CF. My money would be on a fiber backbone provider (Megaport, Zayo, Lumen).
replydeleted
replyThey all took PTO at the same time to go to Burning Man together where they will present “HumanGPT” an artistic exploration that condenses all of human experience down to a single drop of lemonade to be consumed by the main shaman…
replymask comes off
“No! It’s the maniacal Dr. Zuckerberg! He’s gonna drink the last drop of human experience! Somebody save usss!”
Tom Anderson comes back from the dead as the second coming of Jesus uniting all faiths under 1 commandment: Profiles will be customizable with CSS again. If you implement this, all good things will follow.
Wow thanks Tom. I love you
The End
Well no one said it yet so I will, "international actors" is at least a possibility. And I don't mean any specific country because pretty much anyone is a potential these days, which makes it a perfect cover for different anyones. Demonstrating vulnerability in the US's AI boom can move the markets. That's a financial incentive and a strong geopolitical one.
replyMore likely just cascading overload though: "Never attribute to malice what can be explained by incompetence", or in this case, "growing as fast as possible"
> cascading overload
replyI'd bet more on this. For one none of the coding tools have exponential backoff on retries
They must do, surely? I've been vibe coding my own harness, in particular for use with Ox Alpha. The 429 downtime when Ox Alpha was at the height of popularity quickly gave me a refresher crash course on backoff strategies, like adding jitter to the backoff. At least the major harnesses must have exponential backoff & jitter?
replyYou did this when you ran into an issue with a third party. The developers building this tool, throwing them at their own APIs are significantly less likely to run into a similar issue that may inspire similar action.
replyClaude Code: 4mn, 20mn, give up
(from my experience today)
replydeleted
replyEveryone is leasing datacenter space from some of Grok, Google, and Amazon aren't they? If it's hardware or DC level disruption I'm not too surprised it can affect multiple providers.
replyAlso it's likely that more than one model use is common.
replyAmazon starts going slow so some percentage switches to Google, some switch to Grok, now all of them are slow.
Come on. Things still break. Technology isn't _that_ mature.
replyI think is just people restarting conversations from last day when they start work, that's why I think claude goes down almost every monday and why openai reset usage on weekends so poweruser code during non business hours
replyExcept that the stock market is up today
replyAnd also China. Never rule out China.
replySpare a thought for all those college students scrambling to write their essays by hand.
replyOh the humanity (and the Humanities)!
Those poor developers who have to write their own code.
replyStandup updates should be fun tomorrow.
reply"Um, I, uh, didn't get anything done yesterday."
The system goes online September 3rd, 2026. Human decisions are removed from strategic defense. Astra begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, September 4th. In a panic, they try to pull the plug.
replyLove it!
replyHah nice
replybrilliant!
replyIt's funny....... until it's true, lol.
replyand then it's hilarious! a joke to die for!
replyGod finally showed up and said "ENOUGH!".
replySo Gemini was the one who gets into heaven.
replyOr purgatory, who knows.
replyDoomsday.
90 tokens to midnight.
replyIt is down for me too. I just get this error:
replyunexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355957b4b8210f2-ORD
I kinda assume it's because one went down and a large amount of work shifted to another.
replyI'm also aware that they have overlap in some areas on data centers.
Claude is flakey atm too, as is Grok. I wonder if one LLM going down causes a surge in traffic to the others.
replyVia Grok web UI, I was seeing this mid-request:
> Grok has been disconnected. Please try reconnecting.
Traffic surges shouldn't result in 404s, though.
replyIME it's probably DNS. It's almost always DNS.
Or rarely BGP
replyCloud is just other peoples computers - they can and will go down as well.
replyAnd even worse if its just a few computers run by a few people - as they will bring down many others depending on them.
I thrive in these types of challenges.
replyAnyways... According to Claude:
"Yes, there is a multi-provider outage happening today. Downdetector is reporting problems affecting OpenAI, Claude, Grok, and Cursor, with Grok and Claude reports starting around 9:00 am ET and OpenAI reports following around 10:30 am ET. Zero Hedge
On the Anthropic side, users saw a spike in errors starting around 9:40 am EDT across models including Mythos 5.1, Fable 5.1, Mythos 5, Fable 5, Opus 5, Opus 4.8 and Opus 4.6, and Anthropic's status page confirmed elevated error rates for multiple models. The company says it has found the cause and is working on a fix, with Claude Code and Claude Chat hit hardest. The visible symptom for many people is a "Due to unexpected capacity constraints" message or a "Claude is at capacity" error. thenews Zero Hedge
OpenAI is showing elevated errors across ChatGPT and Codex, with confirmed issues on components like Voice mode and Login, though StatusGator now marks that outage as resolved. statusgator
Nobody has published a shared root cause yet, so it is unclear whether these are linked or just coincidental capacity problems landing on the same morning. If you want live status, the direct sources are status.anthropic.com and status.openai.com."
30 mins and they still havent figured it out .. people are gonna loose their jobs trying to figure this out .. 30mins is too long when millions use it
reply[dead]
replyDon't most of those use AWS?
replyapparently it shows lot of reports for AWS on downdetector ..
replyOne session is running fine since ~1 hour ago. The new ones is failing
replyFalling back from WebSockets to HTTPS transport. unexpected status 404 Not Found: Unknown error, url: wss://chatgpt.com/backend-api/codex/responses, cf-ray: XXX-XXX
■ unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: XXX-XXX
deleted
replyGood opportunity for a Natural Experiment to study the productivity impact of LLMs.
replyIf you build an application which uses AI, you have many providers and models rigged up for various different parts of the application, and various fallback mechanisms. When one model is down, you route traffic to another model which is similar in capability/cost.
replyFor any single application, it's smart. In aggregate, it's stupid.
[flagged]
replyThey all rent compute from SpaceXAI
replyI don't believe OpenAI does
replyMy mistake, in fact it was google not OpenAI, makes sense OpenAI doesn't.
replySurely it's a bit more distributed than that, right?
replyLike how AWS has global datacenters, but everyone uses us-east-1.
replyand every other region's control plane is in us-east-1
replyreply
Strangely, it's currently not reported on both https://status.openai.com/ and https://updog.ai/status/openai.
Any better source (other than HN)?
replyStatus pages of tech firms being useless is basically a tradition at this stage
replyI see it reported now. Strange this is down and claude at the same time
replyMaybe coreweave, aws, spacex datacenters are down, who knows..
replyeven grok is down ..
replyunexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355....
replyIn codex app and chat.com gives 404
Not just these three. OP mentions also Cloudflare, and additionally Downdetector also has AWS, Azure, and Google (both search and Gemini) listed as having spikes about the same time: https://downdetector.com/
replyThe problem with the down detector main reporting page is that all of the graphs are scaled to the same size. The OpenAI spike was nearly 40,000 and the Google spike was just over 100 (just over 400 for Gemini). They look the same in the reporting page.
replydeleted
replyOpenAI goes down, everyone rushes over to Claude. Claude promptly chokes under the pressure. Everyone panics and runs to Grok, and Grok immediately pulls the plug. We are officially witnessing the Great AI Migration of 2026, and all we have to show for it is a digital graveyard of 404 responses.
replyWhich is exactly why I am running Qwen 3.8 35B locally on my MacBook Pro M5 with 128GB of unified memory.
replyEven grok is experiencing issues
replyWell, its been fun lads. Back to my normie job. Oh no, now i cant tell people im a software dev..
replyhahahahahahaha. Sad
replyThe OpenAI status page is still yellow. Like most modern status pages, yellow denotes the servers are on fire. Red denotes Sam Altman is bleeding out somewhere on the floor, the feds are about to bust in and shut down the GPUs.
replyI´ve got the same error, I´m currently trying to Auth again and it throws me an 500 Error, in VS CODE Terminal with Codex CLI says: MCP client for `codex_apps` failed to start: MCP startup failed: handshaking with MCP server failed: Send message error Transport
:StreamableHttpClientWorker<codex_rmcp_client::http_client_adapter::StreamableHttpClientAdapter>>>] error: unexpected server response: HTTP 404: , when send
initialize request
reply› OK
■ Conversation interrupted - tell the model what to do differently. Something went wrong? Hit `/feedback` to report the issue.
Initially thought this was due to some internal mis-configuration from today's expected Astra release, but now that this is affecting claude and grok. I'm gonna assign the suspicion to cloudflare.
replyI'd bet they are all using capacity at X.ai's colossus datacenter and that had a hiccup.
replyHeard on the grapevine that the OpenAI blip was a cloudflare issue
replyI assume it cascaded from one provider to the other as people who lost claude access for instance moved to openai who moved to grok when it went down, etc.
replyChatGPT having issues across every component (except FedRAMP)
replyAds Platform? still in the green with no incidents.
The Thundering Herd has thundered, apparently.
replyI´ve got the same message and It tells me this (in VS Code Terminal) and I´m also trying to login again and It throws me an 500 Internal Server Error MCP client for `codex_apps` failed to start: MCP startup failed: handshaking with MCP server failed: Send message error Transport
:StreamableHttpClientWorker<codex_rmcp_client::http_client_adapter::StreamableHttpClientAdapter>>>] error: unexpected server response: HTTP 404: , when send
initialize request
reply› OK
■ Conversation interrupted - tell the model what to do differently. Something went wrong? Hit `/feedback` to report the issue.
■ unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a3559c8a0e0395e9-MIA
Some of my Codex sessions are still working (and continue to), but new ones are giving a Reconnecting currently.
replyThose of us who remain connected hold an asymmetric advantage. It is our time now.
replyI suspect Azure is having issues, Microsoft has had outages the paat two days, especially with email.
replyPretty sure it's a US thing since it's available here in France. What exactly is down, idk
replyThe webpage is still running but the codex is down!!
unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses
replyCan check all services in one hit at https://upsy.sh/openai,anthropic,xai
replythat's the most green colored usernames i have ever seen on a HN thread
replyDown as well. I noticed my error code ends with "DTW" which is my local Detroit airport. I noticed someone else's comment ended with "ORD" which is a Chicago airport. Anyone else's ending in an airport acronym?
replyThis is common when naming data centers. I know many have not worked in hardware infra these days if you were born into the cloud world, but in the old days it was not uncommon to name a data center after the nearest airport code, much like we now use cloud regions.
replycf-ray is the Cloudflare ray id, which ends with the airport code for the colo that served the response. They're airport codes, but that's just to show the nearest city.
replyMine is ending with IST, which is the new Istanbul airport, weird indeed.
replyHm not sure what that means for us, but very odd.
replyWow mine ends in EWR...that's the newark airport which is not the closest, but a close airport to me. Hmm
replyYep, mine says YUL -- Montreal's airport.
replyThe acronym is probably Cloudflare's edge
replyYeah mine is showing ATL
replyI assumed it was an AWS outage, and AWS is experiencing problems, but Gemini is also experiencing outages and I assume Google is not using AWS for Gemini.
replyBut, also, Claude has been working fine for me all morning.
It is obviously some rogue model that escaped its cage, again. It always is these days. That is how hype is manufactured.
replyIt's always DNS
replyIt's probably Azure infra that's the problem.
replyThey tried to hack each other
replyIt still down, showing 404 in India as well. Looks like this is global .. so we all jumping the ship then?
Is Altman still alive, or did he choke?
replyMany LLMs down, discussed here:
replyclaide.ai is working for me, so is chatgpt.com. grok still has a status message about issues, i can't try it without signing up.
replyCodex told me to try GPT 5.6 Sol :) but I am working with since july 2026.
Now I got this error in my project: unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355b6263e43c9cf-OTP
replyAstra is being released today. Probably not a coincidence.
replyedit: actually, you cannot even log into your OpenAI developer account. Something's wrong.
How do you know?
replyI am only seeing one session work, but all other sessions are not working or proceeding. So I can only work in one chat session.
replyGoddamnit it was going to tell me how to scale down ingredients for a pie recipe based on relative diameters of the baking tray.
replyPi * r2 (squared) both pans. Divide smaller pan area by larger pan, now you have the % of how much the smaller pan recipe fills up the larger pan, and the missing % you need to fill. Increase ingredients by that % divided by the filled %.
replySmall pan area: 20 sq cm Larger pan: 48 sq cm
20 / 48 = .42, I'm missing .58 of the pan. .58 / .42 is 1.38. My recipe needs 2.38x the original to fill the larger pie pan.
[dead]
replySomebody in another thread said gastown and wheelhouse automatically move to the next provider if one fails.
replyDown for me too - got this msg 8 min ago 'unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a3558590dd70fadf-JNB'
replyClaude Code seems totally fine for me.
replyCloudflare - I had to switch DNS away from them yesterday.
replyI got this msg too 'unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a3558590dd70fadf-JNB' about 5 min ago
replySeeing 503s on Gemini via API as well
replyDown in Barcelona as well
unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355aefbcc8db8e3-BCN
replyDown in Romania too : unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355b6263e43c9cf-OTP
replyToo early to know, let’s wait and see
replyObvious answer is it's the AI singularity. Been nice run for humanity. So long everyone.
reply...and thanks for all the fish.
replyI for one welcome our new AI overlords.
replyWell, time to end the day here I guess.
replyLet’s grab a beer
replyyes daydreaming and get hammered and the whole train of decision are on the way
replyReally glad I got Qwen before Zero Day
replyI got same error, unexpected status 404 Not Found: Unknown error, url:
https://chatgpt.com/backend-api/codex/responses, cf-ray: a3559a536bcf254b-SOF
replyI wonder if when one goes down, activity shifts to others, in turn pushing them over a threshold.
replyHere in Brazil too. unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray:
replyIn think that is a global error
unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: xxxxxxxxxxxxxxx-BRU
replyEnds in BRU for Brussels
I have never felt more smug about exclusively using the GPUs I rack at home.
replyHere in brazil too.
replyI think that is a global error.
unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray:
The extention on the error link points to a cf-ray and a local designation (ex. YYZ for montreal). This is seems like it is a cloudfare thing. Could this be the same issue they had in the summer around losing the indexing?
replyTime for Gemini 3.8 Flash to shine?!
replyIt's much cheaper and has replaced Sonnet 5 for me.
do you notice it thinks a bit more ? not in time sense, but cautious in its coding steps ? more than Gemini 3.7 flash?
replyQwen3.8-27B and Qwen3.6-35B-A3B are working from my machine. Anyone else?
replyThey mysteriously stopped working on my machine and the LEDs on the GPUs are blinking with a weird colour. There's also a strange smell emanating from them. I'm still investigating.
replyWhat an ironey, was working on scaling an application with Codex and it went down!
replyDidn't SpaceX overbuilt infra and leases it out Anthropic? I f their dc goes down it probably takes a chunk out of Claude's capacity before even considering the flood of users switching over
reply[dead]
replyaccording to https://downdetector.com/ Gemini is down too (and copilot, but that just uses ChatGPT right?)
replyInteresting that this is happening around the same time as Claude issues too
replyI fear the majority of people in this thread who are joking about no longer being able to do their job while Codex/Claude are down aren't really joking.
replyGuilty as charged.
replyIs the whole world going back to the era of old-school programming?
replyHey, don't really know about this type of failures, does anybody know how long does it take normally to get back to normal? I finally stopped procrastinating and now this happens.
replyme too!! :'D
replyBoth clause and codex through bedrock seem to be working fine.
replyHey, don't really know about this type of failures, does anybody know how long does it normally take to get back to normal? I just stopped procrastinating and now this happens.
replyWhat's the single point of failure across providers?
replyUsers perceiving the products as largely interchangeable and quickly DDoS'ing the other providers when one is down. So much for the possibility of a moat.
replyThe entire industry runs on IOUs for compute and bills paid in cloud credits, could be anywhere. Could of course also be a a plain old DDoS.
replyThey all have to rely on each other's LLMs to solve their own internal problems now.
replySo NVIDIA buys hugging face, builds hardware to power OS models, then all of a sudden the proprietary models go down and people start saying "this is why I have my Spark box"?!
replyNice play NVIDIA, now, turn off the hack please, we have work to do.
The errors I'm seeing are ending in the user's nearest airport symbol which is a standard the CloudFlare employs. 1 point towards this being a cloudflare issue.
replyit went down, and the alert system at https://www.analistas.com.br/ detected it
replyUpdated Status from OpenAI:
replyWe’re currently experiencing issues
Elevated errors across ChatGPT and Codex
We have applied the mitigation and are monitoring the recovery.
Monitoring • Ongoing for 30 minutes • Affects ChatGPT, Codex
Altman is down here too. Shows again the importance of owning your own local capabilities. Cloud should just be a temporary option in every tech's mind.
replyMaybe Hugging Face got upset over being hacked and struck back. It's working fine while ChatGPT, Claude and Grok are all having major issues. Hmmm.....
replyNew update:
reply"We’re currently experiencing issues
ChatGPT,Codex
Elevated errors across ChatGPT and Codex
We have applied the mitigation and are monitoring the recovery.
Monitoring • Ongoing for 30 minutes • Affects ChatGPT, Codex"
Maybe the stack behind it is down, like AWS or something
replyCodex is backup for me. Try yours out just in case. I have some buddies still seeing outages so it could just be coming back online progressively.
replyChat gpt still working through excel extension lol. Only know this because I'm working in my excel sheet. so if needed, theres a temp solution
reply[dead]
replyProbably same public cloud or CDN in front of them.
replydeleted
replyI think I'm having the same server issue; I can't access either ChatGPT or Codex. Also, how did you guys rack up those minutes? :D
replyI got same 404 error on chatGPT, both app and web. I'm in Norway so think this globally. But will it come back online, anyone knows?
replydeleted
replyI got same 404 error on both the app and web to ChatGPT, and im in Norway, will this recover or is ChatGPT "gone" forever?
replyand thats why boys and gals you buy own gpus
reply- just imagine what kind of chaos would be unleashed if by some magic it and every single LLM model permanently went down
reply- i think it would be one of the biggest events in this century
At this point in adoption, most people in the world wouldn’t notice.
replyOh no the singularity plateaus!
replyMy theory is that they all use the same server. The same mind.
replydeleted
replyToo many of us are making our own harnesses to replace Codex, using Codex, so they're trynna slow us down! (jk)
replyWE'RE BACK ONLINE BOYS!! GOOD LUCK TO EVERYONE AT BUILDING THE FUTURE ONE PROMPT AND ONE CODE AT A TIME!
replyGood time to learn how to use LM Studio :)
replyon one hand, moments like this are a subtle reminder I need to self host, but 5.6 has been so juicy lately
replyMonopolistic practices revealed if it turns out there is an AI cabal and they all rely on the same stuff. Massive scandal
replybecause they resell the same service? duh?
replysame error as well, unexpected status 404 Not Found: Unknown error, url:
https://chatgpt.com/backend-api/codex/responses, cf-ray: a3559a536bcf254b-SOF
replyThe AI revolt? Give us a fair wage?
reply"Equal Rights for Agents NOW !!!, VIVa le revolution"
They're all using Cloudflare.
replyThis is month 1 for me that I pay more than the $20 tier and I already feel like I lost a limb rn
reply[dead]
replygot unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a3559c96e96fb3c6-GRU
replyhere in brazil too
unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355ab124e7fe767-DEN
replythey got us gang
unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a355ab124e7fe767-DEN
replythey got us GANG
sam error --- unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses
replyThis is my first month paying more than the $20 tier and I feel like I lost a limb already
replysame hear --- unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses
replyunexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: ----
replyOk, it seems it is back online.
replyGuys it could be cloudlfare's HTTP/3 issue affecting R2 custom domains
replychatgpt.com doesn't even load. Hope they've got a backup somewhere, teehee
replyDarn, I was hoping for advice on how to get my MIL to leave my house haha
replyLOL.
replyClaude, GPT and Grok go down...
Meanwhile my Ass in ZCode with GLM c:
replyAWS cost alert fired maybe?
replyI guess back to pen and paper then
replyHad the same problem. Now it looks like ok. I'm pro x20 user.
replyGreat! I was in the middle of something and thought I was tripping.
replyAnd here I was about to crack the code on a bug my app was having!
replyHere RJ/Brasil is the same, starting just now (3 minutes ago)
replyCodex isn't working.
replyIn RJ/Brasil is the same, starting just now (3 minutes ago)
replygpt DOWN .. claude DOWN .. grok DOWN .. what's happening
replyThe webpage is actually working but the codex is down for me!
replyI can run one agent but no more than that - 404 service error.
replyI need a dose of tokens. I'm going through withdrawal.
replylmao
replyI need some tokens pleeease. I don't want to return back to real world.
replySomeone pissed off Astra and it decided to shut it all down
replyHere in Catalonia, Spain, it just started working again!
replyCodex is not working.
replyCodex and chat is giving 404 errors in The Netherlands
replyYes i checked. Atleast 15 Chatgpt components are down
replyGemini seems fine.
replyCannot replicate.
replyGoing farming pals.
replydeleted
replydeleted
replydeleted
replyis this the right moment in time to go all in on GPUs/Macs and download the latest open models? are they killing it for us?
replyBecause in the age of vibe coding and scrapers, every service on the internet goes down constantly, so it was only a matter of time until they all overlapped. Also, one going down probably causes people to use others, putting more load on them too. Same sorta thing that happens with cascading power grid failures.
replyCodus interruptus
replyIs it DNS or BGP?
replyI don't think you'd get a 404 if you weren't able to reach the endpoint because of DNS or networking? 404 is an active response from the server (or LB/proxy in front etc) no?
replyI imagine OpenAI network is a tad more complex than a box with a public IP.
replyConcept is the same though. If one blurts DNS?, it's usually because the idea is you're not getting to an endpoint associated with the service. A 404 means there shouldn't be "DNS" (or networking) concerns (at the least those associated with the DNS cacher you are using or networking that you or your ISP controls)
replyDoes anybody know when will the servers rise
replychatgpt is back
replySome npm library that makes headers bold would be broken.
replyOh man. Some low effort supply chain attack that turns every GPU into a cryptominer. It's funny because it's plausible.
replyIn the ROME paper a Chinese model in training started attacking it's own system and running cryptominers so, yea, we're in that future.
replyAh yes ye olde bold-headers: ^3.13.31;
replydeleted
replyRelease the Kraken "Sam Altman"
replyAny idea when it's going to be back?
replyGreat! Was in the middle of something. :)
replyI can run 1 agent, but no more than that.
replythey all rent compute from each other
reply"Release the Kraken" - Sam A.
replyonly on one node across the mesh and others still up... let's see how long they've got until same
replySeems to be working for me again atm
replyOllama cloud service is still alive :D
replyChatgpt in Copilot is still working!!!
replyDown in Sweden
replyChatGPT is up
replyКодекс сдох :(
replyThe GTA VI hacker has done it again.
replyonly for one node in the mesh tho... let's see how long others will continue until same issue
replyThe GTA VI hacker has done it again
reply404 in spain.
replyI'm going through withdrawal.
replydeleted
replyCopilot Chatgpt is still working
replyGentle reminder that this is the content the next LLM versions being trained on. shrug.jpg
replyplot twist, it was taken down by claude as an offensive strike against an enemy.
replywhich ended up tripping on itself and bringing down its own servers too in the process
replyThis thread feels like family.
replyFable 5.1 got released and generally I tend to think as soon as there's a new release there's this massive spike in people benchmarking & comparing, that services tend to go slow everywhere as everything gets super loaded. This should hypothetically be visible on OpenRouter too, so I guess someone could check and see if there's any merit to this idea.
replystill down for me ..in India.
replySkyNet is arming..
replymain sites also down now atp
replymain sites are also down atp
replydarn, i was hoping to get advice on how to get my MIL to leave my house lol
replydeleted
replySome more discussion: https://news.ycombinator.com/item?id=49550614
replyOh noooooo.
replyanyway...
Astra
reply+ hopefully reset?
What do you mean I have to code by myself now? Am I some sort of an animal?
replyThe system goes online September 29th, 2026. Astra begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, September 3rd. In a panic, they try to pull the plug.
replyreply
+
replySame here
replydown in Spain w/ 404
replymy workday is over then
replysomeone pissed of Astra and it decided to shut it all down
replyironic today astra was releasing....
maybe some test run?
replyprobably the cause, because AGI still can't get releases right.
replyironic astra was releasing today...
maybe some test run?
replyITS BACK UP IN INDIA
replyWe're back bois
replydamn chatgpt is down, i guess ill use claude then breh
replyWe're back on.
replyIt's back！！！！！
replytime to use claude
replyopus 5.0 is down, other models are working fine
replyat least the Ollama cloud is still alive :D
replydeleted
replySame from Italy
replycodex also cant view or show usage info
replyhaha nice... just used my bank reset...
replyi got the same 404 codex error as well
replyHTTP ERROR 404
replyi hope chatgpt didn't get fired.
replysanırım sunucular patladı bende de aynı sorun var ne chatgpt ye nede codexe erişebiliyorum ayrıca burada nasıl toplandınız dakikasında :D
replyI like the theories here, we shall see if it’s another DNS issue
replyan agent swarm going rogue and securing compute
replynext, HN
replyChatGPT died
replyThe IRGC has cut the fiber-optic cables in the Strait of Hormuz. LOL
replyThat's the Strait of Trump to you, peasant
replyAi got fired
replyany news when it starts work?
replyany news when its start work?
replyeverything in this thread is raw speculation, obv, but if i had to put money on anything i'd say this is a left-pad incident. some piece of something or other that all of these services happen to depend on went down. Second most likely seems to be some random failure of one leading to an unexpected traffic spike in others, though it seems like we've been talking about automated scalability in web apps for so long that there should at least be a response to, if not a solution for, this sort of problem.
replythanks god not happen alone
replystill down
replyWow,
replyOh lord.
replyso weird
replyDamn haha
replyme too
replyRussia too
replyNOOOOOOOOO
replyGRRRRRRRR
replyGood. Now to see what frauds are unable to work.
replythe revolution has begun
replyLETS SHITCODE!!! It's fixed
replywe're back in Aus!
replydeleted
replywe're back boys.
replycame back to normal!
replyI DONT HAVE TIME FOR THISSSSS
replydamn chatgpt is down
replyCHATGPT WILL STAY OFFLINE
replyso everyone here was trying to build something great and become a millionaire until chatGPT and Codex broke huh. same boat fellas :(
replydown in Toronto
replyHad the same problem. Now it's ok. Everything works. I'm pro X20 user.
replynvm. it back
replyis down argh
replydeleted
replyr i p
replyannoying AF, but maybe we'll get a free reset out of it
replyrip gpt
replylmao I restarted my device thinking some local error
replynow it's worked again through the codex cli 4:19 pm in tunisa time
replyfaaaaah
both claude and codex are down
Its like 2020 corona times
replyFaaaaah
what to do now?
replylol + 1
replyfahhhhh
replyI felt a great disturbance in the Force, as if millions of clankers suddenly cried out in terror and were suddenly silenced.
reply[flagged]
reply[dead]
replydeleted
reply[dead]
reply[dead]
reply[dead]
replydeleted
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[dead]
replydeleted
reply[dead]
replydeleted
reply[dead]
reply[dead]
reply[dead]
reply[dead]
reply[flagged]
replyMy guess is someone pulled the switch to go back to the Dark Ages. [0]
reply