> I think a lot of people would be better off buying a fast workstation and a very cheap laptop[...]

This!

I'm currently in a hotel room using a cheap laptop (5-year old Lenovo 330) to connect remotely to my fast workstation (home-built Ryzen, lots of memory, lots of storage and LTO6 tape for backups). There's almost 1000mi between them but that's not stopping me doing anything I require. If someone steals the laptop, no big deal, buy another one. All I do is sign and and connect to remote stuff, pretty much nothing of value is stored locally.

"My home [PC] is my cloud", as Sir Edward Coke didn't quite say...

how do you connect to your pc? Through visual code plugin or something?

Tailscale and SSH work great here. I have a ryzen desktop in the office and all ports open in the VPN network, so it is easy to use whatever remote software is needed.

I personally just use vim over ssh.

Not GP but there are a number of remote control software. I'm using Anydesk on windows between laptop and desktop. SSH everywhere else.

Have mentioned this on HN before but I've done this for years, with OpenVPN to connect to my network and then SSH and XPRA for the actual remote X or terminal. On LAN I sometimes use a SPICE VDI. I don't care at all if I break or lose my disposable laptop and there's no data stored on it anyways.

I used to use a desktop workstation for the above, but after I got a rack server (running 24/7 with a bare-metal hypervisor) I'm finding I'm using the desktop less and less unless I require the GPU. If I was back in school now I would've forgone the desktop altogether and gone with a cheap laptop and a used server.

By the way, it's not just about performance. It's about the fact that your data is protected and secure, and you can have backups and redundancy in your private cloud instead of relying on a single laptop. Depending on how your remote access is set up you could even have a browser interface and thus be able to work on a borrowed machine or even on a public library computer.

I don't disagree with your overall sentiment but...

> a very cheap laptop

Overall I think the problem here is just how many things are tied to the price. Quality of screen, keyboard, battery, etc. If it's really only for occasional use, fine, but otherwise, it's probably still worth it to get a mid-ranged but competent laptop to complement the desktop/workstation. (It would be easy to devolve into each user's personal use cases here, on whether to prioritize screen, battery life, etc.)

> RTX GPU ... are just not available in laptops

This is not true, though I'm sure what you mean is "full wattage desktop-version RTX GPUs are not available in laptops." You can get up to a ~150W RTX 3080 in a laptop, plus there are a few "external GPU" solutions like the Asus ROG Flow Z13 which let you plug a desktop-class GPU into your laptop.

> Overall I think the problem here is just how many things are tied to the price. Quality of screen, keyboard, battery, etc.

The secret here is to get an old high end laptop. Then the screen and keyboard are good but the price is low.

That's exactly my approach, too :) Thin client laptop + GPU workstation in the basement.

Docker Image and Proxy are here: https://news.ycombinator.com/item?id=30909178

That's typically how I go. Desktops for heavy lifting, something smaller for occasional portability when needed. In fairness, my current laptop wasn't cheap at around $950 for a then-current i7 processor, 16GB RAM, and nVidia GPU, but it's also 8 years old and still perfectly useful for what I need a laptop to do.

It's not just about saving money either. I just can't really get the performance in a laptop, but I occasionally need to be portable. Not too often, but sometimes. Desktop tends to get the budget and laptop gets the rest.

Have you considered one of those Thunderbolt-based external GPU enclosures? I've never had the money to play around with them, but have always been curious about how to integrate one into a setup. Theoretically, it would let you upgrade the laptop and GPU separately, while also making the GPU more readily available to be reused on some other machine.