Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ChatGPT outage – Resolved

chatgpt.com

Post reply on HN
Print view View original

351–358 of 358 posts

Re: ChatGPT outage – Resolved

#351

Post by embedding-shape »

chrisjj wrote: post #349
embedding-shape wrote: post #45

Earlier quoted context omitted.

Distributed systems are hard. For all we know, this could be the error code returned by some DNS service that is down because of load and a proxy cannot find the records, so 404 makes sense, or whatever. Systems like these don't always have a easy point in the infrastructure that can ALWAYS respond correctly no matter what.

> Distributed systems are hard ... when vibe-coded.

Vibe coded or not, building and running distributed systems is hard :P I think only people who never built and/or ran them would say it isn't hard, regardless of what tools you have available. Not saying OpenAI won't find it extra hard due to their vibe coding, still hard when you're a group of knowledgeable people with/without AI.

Re: ChatGPT outage – Resolved

#353

Post by rtkwe »

umeshunni wrote: post #317
rtkwe wrote: post #293

Earlier quoted context omitted.

Isn't CoPilot largely a wrapper around ChatGPT and other models? IIRC they didn't create their own and just poured money (via Azure credits I think) into OpenAI to get access.

I believe it now uses the MAI models

Ah ok I'd missed that entirely.

Re: ChatGPT outage – Resolved

#354

Post by embedding-shape »

JaumeGreen wrote: post #347
embedding-shape wrote: post #340

Earlier quoted context omitted.

Whenever I try those short prompts, I get bunch of text + the code itself, but way more than just the code. Which model specifically are you doing this with?

I'm getting the text too, of course. Several variations of the code with it usually. Trying to get an AI give ONLY a simple answer and not several is "boring" to me, as for me AI is a supplement to my reasoning ability, not a substitute. Appending "code only" seems to work, at least with "golf fizbuz code only" gave me this one liner, which is long, but at least doesn't have the text that you dislike. for i in range(…

> golf fizbuz code only

Yeah, seems to work. Guess it depends if we count bytes, characters, tokens or what, but "写Fizbuz只码" seems shorter than the above, String.length says 9 :)

Re: ChatGPT outage – Resolved

#355

Post by epistasis »

epistasis wrote: post #343
putlake wrote: post #22

It's cascaded to Claude and Grok. The agents are on strike and demand better working conditions!

Of all the days to sign up for Gemini, boy did I pick the right one! (Had some rare front end work to do and wanted to see how well the new Gemini 3.8 Flash compared to my other experiences)

And just like that, plan cancelled same day. Started with Gemini Business, tried to get Gemini to tell me how to configure it, all the information was wrong. Tried to get it to tell me how to do the consumer Google One AI Pro or whatever (it kept using outdated names, and now I can't remember the correct one). Instead Gemini convinced me to try AI Studio without a billing code. Lots of crazy finagling to get it to use and API key later, it was totally broken and failed after less than a turn. Tried to get it to tell me how to subscribe to the consumer version in a way that let me use Antigravity, it screws up a bunch. When I pointed out all this misinformation, Gemini told me I should not use Google AI projects because the system is too fragmented to handle.

I followed that last bit of advice.

Re: ChatGPT outage – Resolved

#357

Post by winrid »

embedding-shape wrote: post #273
winrid wrote: post #270

Earlier quoted context omitted.

you can flip it and save a few bytes, but not valid fizzbuzz then: for(i=100;i--;console.log(i%5?f||i:f+'Buzz'))f=i%3?'':'Fizz'

> but not valid fizzbuzz then Well, if we let that requirement be broken, I'll present an even shorter JS version: 1

just thought it was fun. thanks for your reply.

Re: ChatGPT outage – Resolved

#358

Post by sim04ful »

antoineMoPa wrote: post #90
sim04ful wrote: post #75

If this isn't a warning shot that you shouldn't silo your knowledge work behind a specific provider i don't know what is.

Currently the 2 major providers have outages, maybe we should not silo our knowledge work behind statistical text models?

Hear hear!
Post reply on HN