#From the nextjs community on Reddit
wise burrow
I wonder if this has more to do with flexibility of scaling hosting infrastructure more than anything (getting off of Vercel $$$).
People are talking about making ChatGPT more CSR-forward, but isn't Next.js still perfectly capable of leaning into static pages and CSR?
delicate thicket
There have been many discussions on this. Theres no way to know the answer without someone from the organization telling why. It could have been as simple as a new lead dev getting hired and saying 'I like remix lets move to that' Theres too many variables. People are the most complex variable of all, not everything is driven by money, someone could have simply said, remix is better for client side we dont need any ssr, boss says okay.
The front end move from nextjs to remix is so minimal for the chatgpt frontend its literally pointless to guess why, the front end is simple they could use zero framework and pure react without any downside.
delicate thicket
I wonder if this has more to do with flexibility of scaling hosting infrastructu...
Who is to say they were ever actually hosted on vercel? I host an internal application on internal hardware, get 10-20k users a day, its not hard to scale especially when you dont NEED vercels backend in any way.
wise burrow
Who is to say they were ever actually hosted on vercel? I host an internal appli...
Yeah, our team self-hosts our Next.js app as well and it works great, but we have a simple setup. I wonder if it becomes more of a headache when you want to scale to multiple servers or an edge network.
Wes Bos said that they used to be on Vercel in his video breakdown (https://www.youtube.com/watch?v=hHWgGfZpk00) and I just took his word for it.
I'm also curious whether ChatGPT was using App Router or Pages Router. Idk if anyone knows.
delicate thicket
He don’t know shit lol. We have 4 instances of our nextjs app internally with a load balancer and a separated authentication process it’s pretty easy to scale.
Edge is a different story.
Haven’t had the need to work with it.
And yeah no way to know if they were on app router or pages.
grim warren
Yeah, our team self-hosts our Next.js app as well and it works great, but we hav...
Actually scaling on the edge will cost a ton of money
and wdym by multiple servers?
the edge handles that for u, u just need to provide your src and the money
And also always sleep with an eye open cause the bill can explode in seconds
for whatever reason
red osprey
**Actually** scaling on the edge will cost a ton of money
Umm, Ton of money? $0.5/Million reqs.
delicate thicket
I’m sure ChatGPT generates enough to not sleep with any eyes open
grim warren
I’m sure ChatGPT generates enough to not sleep with any eyes open
Im not talkin about chatgpt
grim warren
Umm, Ton of money? $0.5/Million reqs.
Yeah, for ur personal stuff...
u aint reaching more then 50 users simultaneously...
red osprey
Yeah, for ur personal stuff...
wdym ur personal stuff? I'm talking about a production app
delicate thicket
This whole thread is about chatgpt.
red osprey
u aint reaching more then 50 users simultaneously...
I'm talking about pay as you go
grim warren
This whole thread is about chatgpt.
Very chatgpt like
wise burrow
**Actually** scaling on the edge will cost a ton of money
costs a ton of money on Vercel, right?
But what if you are OpenAI and can switch to scaling to edge with Azure instead?
That's what I am getting at
grim warren
i just answered to that :)
delicate thicket
costs a ton of money on Vercel, right?
But what if you are OpenAI and can switch...
This is likely precisely what they are doing.
but again, no way to know anything, they could be hosting frontend on their own hardware, really no way to know. Not even worth really hypothesizing about. Microsoft could providing distributed hardware.
grim warren
costs a ton of money on Vercel, right?
But what if you are OpenAI and can switch...
They're probably hosting on their own infrastructure, tbh. But isn't Azure also expensive? I'm pretty sure they don't want to lose any of their profit
delicate thicket
It cost less than Vercel.
red osprey
Cf pages is the cheapest. I mean I've estimated approx $0.5/mil reqs.
wise burrow
They're probably hosting on their own infrastructure, tbh. But isn't Azure also ...
iirc they have a ton of Azure credit from Microsoft
delicate thicket
Microsoft invested in openai, they likely get much better deals
grim warren
iirc they have a ton of Azure credit from Microsoft
Yeah
delicate thicket
We get non-normal prices working with Amazon, we are partnered and have custom Echo hardware/software in the hospitals having a partnership with Amazon changes ALOT of things.
especially since we are the first facility of our kind to integrate alexa into patient facing applications and uses.
grim warren
especially since we are the first facility of our kind to integrate alexa into p...
Alexa + hospital?
sounds weird