503 errors and degredated performance

<img width="1394" height="123" alt="Image" src="https://github.com/user-attachments/assets/dfba64d0-0186-4267-828a-053fe47eeb5d" />

Same here. Not again, please... I need to use it for work. So glad I cancelled my private max subscription.

I had issues this morning too in the desktop app, it looks like they were rolling out an update as the UI changed shortly after but wasn't great as I had some important work to deliver.

Still hanging on with my subscription but they really do need to address performance some and compensate accordingly :(

Got this issue after updating to the latest version.

  ⎿  API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection termination) · Retrying in 4 seconds… (attempt 4/10)
  ⎿  API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection termination) · Retrying in 10 seconds… (attempt 5/10)
  ⎿  API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 17 seconds… (attempt 6/10)
  ⎿  API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection timeout) · Retrying in 38 seconds… (attempt 7/10)
  ⎿  API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection timeout) · Retrying in 35 seconds… (attempt 8/10)

On average it took 6 to 8 failures like above before it can complete 1 request. (and then repeatr)

<img width="912" height="171" alt="Image" src="https://github.com/user-attachments/assets/bf2c758d-568f-4fb6-817e-7bae69cda389" />

Also, forced logged on and can't log back in to Claude.ai

<img width="488" height="234" alt="Image" src="https://github.com/user-attachments/assets/75406136-9609-4d6a-8ad5-b1f2a9afff79" />

<img width="3456" height="362" alt="Image" src="https://github.com/user-attachments/assets/9be1226f-8965-4ba6-b30d-95a776b2b0bd" />
Same here

Same here

Same here

<img width="801" height="139" alt="Image" src="https://github.com/user-attachments/assets/f83d7d85-58d2-4d68-9be0-afa5c66e631e" />

It doesn't work in all sessions

@humanwritten

Still hanging on with my subscription but they really do need to address performance some and compensate accordingly :(

My subscription will run out this weekend. I am trying out opencode and AWS bedrock. I find it to be really promising so far. Subjectively better I would say.

just got a notification the triage bot failed too looks like its totally down

<img width="491" height="388" alt="Image" src="https://github.com/user-attachments/assets/3270504a-8149-4e04-a693-1cf3d7700ad0" />

@humanwritten > Still hanging on with my subscription but they really do need to address performance some and compensate accordingly :( My subscription will run out this weekend. I am trying out opencode and AWS bedrock. I find it to be really promising so far. Subjectively better I would say.

I need to test out bedrock + some other tools to have a fallback / maybe some promising alternatives.

⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 5 seconds… (attempt 4/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection timeout) · Retrying in 8 seconds… (attempt 5/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection timeout) · Retrying in 18 seconds… (attempt 6/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 38 seconds… (attempt 7/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection termination) · Retrying in 32 seconds… (attempt 8/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 36 seconds… (attempt 9/10)

same issue

<img width="1888" height="336" alt="Image" src="https://github.com/user-attachments/assets/53420454-7987-41b1-9234-f7966bfe6083" />

same issue :(

<img width="1906" height="532" alt="Image" src="https://github.com/user-attachments/assets/4b311a17-6d57-4e47-aff0-c431e7ca7821" />

so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on)

> @humanwritten > > Still hanging on with my subscription but they really do need to address performance some and compensate accordingly :( > > > My subscription will run out this weekend. I am trying out opencode and AWS bedrock. I find it to be really promising so far. Subjectively better I would say. I need to test out bedrock + some alternatives to have a fallback / maybe some promising alternatives.

bedrock is much faster. You pay per token though. But you have the choice to strike a good price to performance balance. I will also try connecting it to LM Studio this weekend. I have MacBook with a decent amount of memory. I will try running it with 30B qwen 3 coder MLX and some similar models. I'll see how that goes.

Also running into the same issue:

<img width="1837" height="146" alt="Image" src="https://github.com/user-attachments/assets/1ab37f4c-b0c4-4bd5-a525-6376f69299d1" />

Claude Code v1.0.110
L Session ID: 98310cb4-c301-485a-b96d-0c7d1c9d2d65

I don’t understand what I’m subscribing to 20x max for.

⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 5 seconds… (attempt 4/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection timeout) · Retrying in 8 seconds… (attempt 5/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection timeout) · Retrying in 18 seconds… (attempt 6/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 38 seconds… (attempt 7/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: connection termination) · Retrying in 32 seconds… (attempt 8/10)
⎿ API Error (503 upstream connect error or disconnect/reset before headers. reset reason: remote connection failure, transport failure reason: delayed connect error: Connection refused) · Retrying in 36 seconds… (attempt 9/10)

Here we go again guys...

Desktop app is dead also

<img width="1112" height="912" alt="Image" src="https://github.com/user-attachments/assets/96fd7b66-b7d4-49eb-bdd2-9c3989cb6200" />

I don’t understand what I’m subscribing to 20x max for.

For the experience.

same mr claude is not working

<img width="542" height="172" alt="Image" src="https://github.com/user-attachments/assets/c38bb463-e4f6-4675-83e4-c7814ff78e19" /> uhh this doesn't look good

I just downgraded to 5x. I will spend rest of 118 on other stuff. Can buy like 5-6 other

And now 20%+ of the software engineering community is twiddling their thumbs.

I don’t understand what I’m subscribing to 20x max for.

@ryoya1122 when you have the answer, tell me too.

<img width="938" height="170" alt="Image" src="https://github.com/user-attachments/assets/4a0281cd-5f85-4357-885e-f22cbaf5a2e0" />

At least this time they do not take 4 hrs to notice. Yaaaay, progress.

so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on)

My prod is on fire and I have no idea to fix it without the help of Claude 🤣

> so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on) My prod is on fire and I have no idea to fix it without the help of Claude 🤣

@linux-root restart the server to see if it works :)

> so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on) My prod is on fire and I have no idea to fix it without the help of Claude 🤣

You have something in common with the Anthropics engineers, I guess 🤣

> so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on) My prod is on fire and I have no idea to fix it without the help of Claude 🤣

try codex from OpenAI... its another terminal based agent

And now 20%+ of the software engineering community is twiddling their thumbs.

do you mind if I quote this - let me know if you want anon or not :)

We should also reflect on how much we’ve been relying on AI in our lives and work.

> > so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on) > > > My prod is on fire and I have no idea to fix it without the help of Claude 🤣 You have something in common with the Anthropics engineers, I guess 🤣

do you mind if I quote this - let me know if you want anon or not :)

> > so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on) > > > My prod is on fire and I have no idea to fix it without the help of Claude 🤣 You have something in common with the Anthropics engineers, I guess 🤣

LOL. They might need to switch to Cursor 🙃

It's working for me now

<img width="490" height="833" alt="Image" src="https://github.com/user-attachments/assets/048473ab-e7aa-4225-8e73-21bf44b5b1f4" />

> > > so while we wait for our anthropic to turn it off and back on again what's everyone working on? :) (or was working on) > > > > > > My prod is on fire and I have no idea to fix it without the help of Claude 🤣 > > > You have something in common with the Anthropics engineers, I guess 🤣 do you mind if I quote this - let me know if you want anon or not :)

Go ahead, quote me properly, though. It's quite fine work. I surprised myself :D

working for me now

We should also reflect on how much we’ve been relying on AI in our lives and work.

we really should all jump on a twitter space or something when claude code goes down and reflect on this 🤣

Relax guys, they already know and working on restoring services Im sure. Try journaling or drawing out your ideas instead until further notice. 9AM fires suck.

> We should also reflect on how much we’ve been relying on AI in our lives and work. we really should all jump on a twitter space or something when claude code goes down and reflect on this 🤣

I like the idea of this, the Claude waiting room

> We should also reflect on how much we’ve been relying on AI in our lives and work. we really should all jump on a twitter space or something when claude code goes down and reflect on this 🤣

lol I hate writing a single line now, I would rather spend 5 minutes telling claude what to do

btw it's back up guys

<img width="717" height="348" alt="Image" src="https://github.com/user-attachments/assets/f82fd9e7-aceb-41b3-b4ee-fa878d83bd7c" />

> > We should also reflect on how much we’ve been relying on AI in our lives and work. > > > we really should all jump on a twitter space or something when claude code goes down and reflect on this 🤣 lol I hate writing a single line now, I would rather spend 5 minutes telling claude what to do

it's so pointless to write code when claude can write and debug it 100x faster than I can on my best day.

Did they spin up a dumb version?

I define my requirements and it just ignores most of them. Simple and comprehensive task from a clean slate with examples given, which it ignores.

Good thing they now give themselves 5 years to debug these issues. It seems like they need it.

resolved here...

Did they spin up a dumb version? I define my requirements and it just ignores most of them. Simple and comprehensive task from a clean slate with examples given, which it ignores.

explicitly tell claude to read the files in FULL, if you're following along you'll see it will often only read a 50-100 lines of a 800 line documentation file.

Did they spin up a dumb version? I define my requirements and it just ignores most of them. Simple and comprehensive task from a clean slate with examples given, which it ignores.

Who knows... Might have deployed further quantised models for subscription folks

There was an incident this morning. It should be resolved now, let us know if you're seeing anything weird.

https://status.anthropic.com/incidents/k6gkm2b8cjk9

There was an incident this morning. It should be resolved now, let us know if you're seeing anything weird. https://status.anthropic.com/incidents/k6gkm2b8cjk9

Looking good to me now, thanks for sorting :)

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.