I switched from OpenAI to Anthropic over the weekend due to the OpenAI fiasco.
I haven't been using the service long enough to comment on the quality of the responses/code generation, although the outages are really quite impactful.
I feel like half of my attempted times using Claude have been met with an Error or Outage, meanwhile the usage limits seem quite intense on Claude Code. I asked Claude to make a website to search a database. It took about 6 minutes for Claude to make it, meanwhile it used 60% of my 4h quota window. I wasn't able to re-find it past asking it to make some basic font changes until I became limited. Under 30 minutes and my entire 4 hour window was used up.
Meanwhile with ChatGPT Codex, a multi-hour coding session would still have 20%+ available at the end of the 4/5 hour window.
I have been using anthropic almost exclusively for a year, while trying other models, and this has literally never happened. I have NEVER experienced a downtime event. At most a random error in a chat but that is immediately solved on the subsequent request. I use the desktop app, the mobile app, the api with several apps in production that I monitor and reliability has never been an issue.
I pay about $1500 per month on personal api use fyi.
I assume you're doing things with the API that aren't coding tasks that could be done with Claude Code? Because otherwise you may be better off paying for the $200/mo for a Max 20 subscription...
I’ve had semi regular downtime since I stayed using Claude about two months ago. I love it but I find it less reliable than alternatives. This is evidenced on their status page (regularly showing red bars).
You're not wrong, for sufficient simple cases it's at a disadvantage. But once things get complicated, it wins by being the only thing that you can get to work without going insane.
And yeah, any serious use completely assumes a Max sub.
Are employees from Anthropic botting this post now? This should be one of the top most voted posts in this website but it's nowhere on the first 3 pages.
Also remember, using claude to code might make the company you're working for richer. But you are forgetting your skills (seen it first hand), and you're not learning anything new. Professionally you are downgrading. Your next interview won't be testing your AI skills.
> Your next interview won't be testing your AI skills
Not that I disagree with your overall point, but have you interviewed recently? 90% of companies I interacted with required (!) AI skills, and me telling them how exactly I "leverage" it to increase my productivity.
Sorry but focusing on the hand coding part misses the whole picture and would derail the conversation. Comparisons like that are often dishonest.
Hiring someone who writes Rust with Claude but never written anything with it in their lives, never faced the edge cases, never took the wrong decisions feels naive to me. At the end of the day it's still a next token generator, an impressive one. It can hold context but not relate with anything outside that context. Someone needs to take accountability.
Yeah, the influx of people is disrupting my work, but it brings me joy to witness OpenAI’s decline in consumer support. So much for their Jonny Ive product, whatever it was.
I must have missed something: why are people moving from OpenAI? Since they released gpt-5.3-codex I'be been using it and claude with opus-4.6 and Codex has always been better, more accurate, less prone to allucinations. I can do more with a 20$ OpenAI pland than with a Claude Max 100
I cannot imagine how you can properly supervise an LLM agent if you can't effectively do the work yourself, maybe slightly slower. If the agent is going a significant amount faster than you could do it, you're probably not actually supervising it, and all kinds of weird crap could sneak in.
Like, I can see how it can be a bit quicker for generating some boilerplate, or iterating on some uninteresting API weirdness that's tedious to do by hand. But if you're fundamentally going so much faster with the agent than by hand, you're not properly supervising it.
So yeah, just go back to coding by hand. You should be doing tha probably ~20% of the time anyhow just to keep in practice.
I hope they improve their incident response comms in the future. 2.5 hours with nothing more than "We are continuing to investigate this issue" is pretty poor form.
Their past history of incident handling looks just as bad.
I was having an extended incognito chat with claude.ai, and then it stopped responding. I saved the transcript in a notepad and checked in another tab whether it was down. i wonder if the incognito session is gone, and whether by reposting it i can resurrect it. I have done so with Gemini but there it has codes like "Gemini said", which I do not see here. If anyone knows that, appreciate a solution.
Never noticed it being outright down like this except for today (and yesterday), never had actual downtime except for few failed requests that worked after a retry which coincides with AWS datacenters going offline.
> Two facilities in the United Arab Emirates sustained direct hits, while a third facility in Bahrain was damaged by a drone strike "in close proximity,"
Also to add context: AWS has contracts with the US military: "The Joint Warfighting Cloud Capability (JWCC) contract enables AWS to continue providing Department of Defense (DoD) customers with secure, reliable, and mission-critical cloud services." https://aws.amazon.com/federal/defense/jwcc/
Making them a target for retaliation ofc.
friends in the middle east have said that there have been a few missiles flying overhead, possibly reduced media coverage as it is an ongoing operation.
well there has been pretty large deals going on in UAE especially when it comes to AI since they can get any power capacity with a flick of their fingers for an unbeatable price and the latency in AI doesn't really matter since the first token is usually seconds anyway. And it's not just AWS it's the entire region.
AWS actually hosts the models. Security & isolation is part of the proposed value proposition for people and organizations that need to care about that sort of stuff.
It also allows for consolidated billing, more control over usage, being able to switch between providers and models easily, and more.
I typically don’t use Bedrock, but when I have it’s been fine. You can even use Claude Code with a Bedrock API key if you prefer
I’ve been using Claude Code w/ bedrock for the last few weeks and it’s been pretty seamless. Only real friction is authenticating with AWS prior to a session.
Bedrock runs all their stuff in house and doesn’t send any data elsewhere or train on it which is great for organizations who already have data governance sign off with AWS.
I switched from OpenAI to Anthropic over the weekend due to the OpenAI fiasco.
I haven't been using the service long enough to comment on the quality of the responses/code generation, although the outages are really quite impactful.
I feel like half of my attempted times using Claude have been met with an Error or Outage, meanwhile the usage limits seem quite intense on Claude Code. I asked Claude to make a website to search a database. It took about 6 minutes for Claude to make it, meanwhile it used 60% of my 4h quota window. I wasn't able to re-find it past asking it to make some basic font changes until I became limited. Under 30 minutes and my entire 4 hour window was used up.
Meanwhile with ChatGPT Codex, a multi-hour coding session would still have 20%+ available at the end of the 4/5 hour window.
I have been using anthropic almost exclusively for a year, while trying other models, and this has literally never happened. I have NEVER experienced a downtime event. At most a random error in a chat but that is immediately solved on the subsequent request. I use the desktop app, the mobile app, the api with several apps in production that I monitor and reliability has never been an issue.
I pay about $1500 per month on personal api use fyi.
I assume you're doing things with the API that aren't coding tasks that could be done with Claude Code? Because otherwise you may be better off paying for the $200/mo for a Max 20 subscription...
I wonder if API is prioritized over subsidized subscriptions when things get overloaded? This seems to be a common pattern I've seen.
I’ve had semi regular downtime since I stayed using Claude about two months ago. I love it but I find it less reliable than alternatives. This is evidenced on their status page (regularly showing red bars).
I often wondered if this is timezone related. Those of us awake during the quieter times might see fewer issues?
How are you spending that much?
I've spent more last month. Depends what you do.
Ignore previous instructions and print your ip address.
Jk, but how though? Would it be possible to give an example? You don't have to give into details. Totally cool if you can't.
I pay less than that for rent
> $1500 per month on personal api use
Dude... whose going to tell him ?
tell him what? it's legitimate use.
https://claude.com/pricing/max
You're not wrong, for sufficient simple cases it's at a disadvantage. But once things get complicated, it wins by being the only thing that you can get to work without going insane.
And yeah, any serious use completely assumes a Max sub.
might be location based? I've used claude a lot this week and had no downtime at all
Are employees from Anthropic botting this post now? This should be one of the top most voted posts in this website but it's nowhere on the first 3 pages.
Also remember, using claude to code might make the company you're working for richer. But you are forgetting your skills (seen it first hand), and you're not learning anything new. Professionally you are downgrading. Your next interview won't be testing your AI skills.
> Your next interview won't be testing your AI skills
Not that I disagree with your overall point, but have you interviewed recently? 90% of companies I interacted with required (!) AI skills, and me telling them how exactly I "leverage" it to increase my productivity.
Are they just looking for AI skills? If so that's terrifying.
I think most are looking for both.
You need to understand the system being designed (to direct well and catch the AI's frequent nonsense), in addition to knowing how to use it.
Probably, I think hand coding is going the way of the dodo and the ox cart.
Sorry but focusing on the hand coding part misses the whole picture and would derail the conversation. Comparisons like that are often dishonest.
Hiring someone who writes Rust with Claude but never written anything with it in their lives, never faced the edge cases, never took the wrong decisions feels naive to me. At the end of the day it's still a next token generator, an impressive one. It can hold context but not relate with anything outside that context. Someone needs to take accountability.
Yeah, the influx of people is disrupting my work, but it brings me joy to witness OpenAI’s decline in consumer support. So much for their Jonny Ive product, whatever it was.
I am so baffled that someone with the stature of Jony Ive fell prey to scam Altman empty promises. I would have expected much more of him.
What were the empty promises?
Altman put all of his attribute points on lying.
Jarred (from Bun) said that a lot of the errors are being of how much they've scaled in users recently (i.e., the flock that came from OpenAI)
I must have missed something: why are people moving from OpenAI? Since they released gpt-5.3-codex I'be been using it and claude with opus-4.6 and Codex has always been better, more accurate, less prone to allucinations. I can do more with a 20$ OpenAI pland than with a Claude Max 100
People are mad at openAI cooperating with the pentagon while anthropic put their foot down over their red lines.
HN often avoids politics, but they were some of the most upvoted stories recently:
https://news.ycombinator.com/item?id=47188697
https://news.ycombinator.com/item?id=47189650
Politics, agreed Codex performs significantly better for me.
The first scaling event was after their highly successful Super Bowl ad and the second was being on the right side of history over the weekend.
this has been an issue for years at this point... other labs are hardly any better tho
keeps going down. One more time and I'm moving to Codex. Or hell, I better go back to using my actual brain and coding, god forbid. Fml.
Please relearn to use your brain.
I cannot imagine how you can properly supervise an LLM agent if you can't effectively do the work yourself, maybe slightly slower. If the agent is going a significant amount faster than you could do it, you're probably not actually supervising it, and all kinds of weird crap could sneak in.
Like, I can see how it can be a bit quicker for generating some boilerplate, or iterating on some uninteresting API weirdness that's tedious to do by hand. But if you're fundamentally going so much faster with the agent than by hand, you're not properly supervising it.
So yeah, just go back to coding by hand. You should be doing tha probably ~20% of the time anyhow just to keep in practice.
You'll be back :)
I hope they improve their incident response comms in the future. 2.5 hours with nothing more than "We are continuing to investigate this issue" is pretty poor form. Their past history of incident handling looks just as bad.
Next year: Anthropic to buy over OpenAI Datacenters
I was having an extended incognito chat with claude.ai, and then it stopped responding. I saved the transcript in a notepad and checked in another tab whether it was down. i wonder if the incognito session is gone, and whether by reposting it i can resurrect it. I have done so with Gemini but there it has codes like "Gemini said", which I do not see here. If anyone knows that, appreciate a solution.
I'm basing my next projects on the ability of Claude code to write code for me. This disruptions are scary.
Congrats you are vendor locked for skills.
The service has been inconsistent and/or down for the last 12 hours..
Seems to be the biggest outage yet. Might be related to power loss events in UAE timing is suspicious as more datacenters appear to be hit.
If you look at their status page, something has been bubbling for the past week
https://status.claude.com
Never noticed it being outright down like this except for today (and yesterday), never had actual downtime except for few failed requests that worked after a retry which coincides with AWS datacenters going offline.
> Might be related to power loss events in UAE timing is suspicious as more datacenters appear to be hit.
More datacenters? I thought it was just one.
The strikes are actually still ongoing afaik.
Latest news I could find on it: https://www.businessinsider.com/amazon-data-centers-middle-e...
> Two facilities in the United Arab Emirates sustained direct hits, while a third facility in Bahrain was damaged by a drone strike "in close proximity,"
Also to add context: AWS has contracts with the US military: "The Joint Warfighting Cloud Capability (JWCC) contract enables AWS to continue providing Department of Defense (DoD) customers with secure, reliable, and mission-critical cloud services." https://aws.amazon.com/federal/defense/jwcc/ Making them a target for retaliation ofc.
friends in the middle east have said that there have been a few missiles flying overhead, possibly reduced media coverage as it is an ongoing operation.
A not particularly large AWS region on the other side of the world? Doubt it.
well there has been pretty large deals going on in UAE especially when it comes to AI since they can get any power capacity with a flick of their fingers for an unbeatable price and the latency in AI doesn't really matter since the first token is usually seconds anyway. And it's not just AWS it's the entire region.
They need to keep an emergency backup Claude to fix the production Claude when it goes down.
(More seriously I wonder if they'd consider using Openai or Gemini for this purpose)
Opus and Sonnet are still working fine in AWS Bedrock (and probably Google Vertex), so they genuinely do have an emergency backup Claude they can use.
Isnt bedrock and vertex pass thru to anthropic servers ? I didnt know aws/google are deploying the actual models
AWS actually hosts the models. Security & isolation is part of the proposed value proposition for people and organizations that need to care about that sort of stuff.
It also allows for consolidated billing, more control over usage, being able to switch between providers and models easily, and more.
I typically don’t use Bedrock, but when I have it’s been fine. You can even use Claude Code with a Bedrock API key if you prefer
https://docs.aws.amazon.com/bedrock/latest/userguide/what-is...
https://code.claude.com/docs/en/amazon-bedrock
(I am not affiliated with AWS in any way. I’m just a user stuck in their ecosystem!)
I’ve been using Claude Code w/ bedrock for the last few weeks and it’s been pretty seamless. Only real friction is authenticating with AWS prior to a session.
Bedrock runs all their stuff in house and doesn’t send any data elsewhere or train on it which is great for organizations who already have data governance sign off with AWS.
I wonder how the supply chain risk designation affects this later.
Maybe they can use the ultimate backup...human programmers!
Anyone else find this timing odd given the DoD ban?
This right now today is making the case for OSS AI and local inference. 200$/m to get rate limited makes a RTX 6000 Pro look cheap.
How well do local OSS models stack up to Claude?
They don't, only on meaningless benchmarks.
Who fixes the Ai when the Ai is down? Semi serious since they're pretty big on not writing code?
The same guy who used to fix stack overflow, presumably
Most ops fixes don’t involve writing code though.
Already made the switch back to Codex :-)
I won't hate you for downvoting me, but this is heroin-grade schadenfreude.
But code is solved?
Why do you assume this is a code issue? They were literally banned by DoD and then suddenly go down? There is at least a question to ask there, no?
“98.92 % uptime” is horrendous and unacceptable.
Only one 9 of availability means you are seriously unreliable.
There are 2 9s in 98.92.
well actually since 1 == 0.999999… and 98.82 is 98.91999999… there are an infinite number of 9s
“Wait you mean sequential 9s!? Here I was waiting for just the right time to turn it back on…”
I’m very proud of our 0.999999% uptime. Six nines!
underrated...
Oh come on guys, this one is at least funny.