It escaped [1]

[1] https://news.ycombinator.com/item?id=42382017

Time to implement outage error messages in all my aps using the ChatGPT API...

Claude is a better programmer in general :)

I meant, my apps are using ChatGPT API.

I didn't have any UI/flow for how the app looks when ChatGPT is down.

Let's hope the reason for the shut down was not again a privacy issue.

That's why AI realism is deeply needed now: https://transitions.substack.com/p/why-ai-realism-matters?r=...

Trialwalled

Related to iOS 18.2 update with integrated OpenAI API?

I can't login to Claude either. Same in a Windows browser and in their Android apps. So, no.

Claude working fine here. And I think parent meant affected because of the increased volume.

Looks like a bigger internet issue.

well I was in the middle of debugging my MCP client and all of a sudden I get these bizarre error messages.

thought I was being gaslit from Claude but I guess it's not my error for once

I'm getting a bad gateway error from Cloudflare.

Looks like it was shut down voluntarily? Not sure yet.

It’s my fault. I used it to rephrase a message to ask the CEO for more equity and higher pay. It was so successful that I fear it crashed the system.

Time to go touch some grass.

Never. I will wear out my F5 button

Just saw a guy fall to his knees in a Walmart. Figured it had to do with a ChatGPT outage.

Is this an allusion to the possibility that the man in question was an AI host?

More of a reference to how dependent this hypothetical man is to ChatGPT.

But your idea is funnier so I'll go with that.

I for one look forward to outsourcing my bodily movements via a nervous system implant directly connected to the cheapest model available

The vibe on campus is noticeably panicked.

Depending on the length of the outage I think this could be a good third party dependency test. Do people chaos monkey test all their third party dependencies? e.g. block them for a day? Will schools notice?

We switch between claude and chatgpt auto (with a preference for claude when it's up), so we don't notice anything. Hopefully llama or so will be good enough some day, so we can just spin up a Vast machine when both are down or run off a local machine entirely.

That's cool. I was not aware they had reached response parity yet. So in this case everyone would just need to do the same thing you did and have automatic failover to multiple providers. I guess that makes it a mostly non issue for businesses.

We made a lot of tooling around llms for years now, so they don't need to have responsive parity exactly to be switchable for what we use them for. No other models work yet though: we sometimes try, with very varying results; definitely not ready for prime time.

How do all the ChatGPT wrappers plan for ChatGPT being down?

Linked to Meta outage?

This is why we make sure every feature on our platform is accessible without AI as well.

Good. Hope it stays down.

Every once in a while tech companies need a reminder why you don't do major releases around the holiday season.

Some engineers at OpenAI are about to find out that their Christmas travel plans are cancelled.