2.5’ hdd but besides running tight on space

No wonder - that’s huge! (sorry - couldn’t resist)

While I don’t exactly intend to run RAID

At these sizes you may want to consider it - at least a mirror (RAID1). That’s a lot of data to lose and/or have to fetch from backups. Being able to limp along until you get a replacement is an enormous time-saver.

Underrated comment.

A “howto” that just gives you scripts and commands to run is pointless. You need to understand the technology, and networking in particular.

No?

2Gig on S3 (in us-east-1) is like $0.05/mo. and is enough for either Nextcloud or Immich. Your data is going to be the largest consumer of space.

Things like Immich or Nextcloud are far too much for what I need, I basically just need a password-protected upload interface and the ability to grab the direct links to the images to embed them.

Why do you care that they do things you don’t need if they also do what you need?

Because these do just what you need and do it well.

Reverse proxies do not give you security.

Agree - critical infrastructure should have as few dependencies as possible.

So - I setup that model according to the docs and gave it this prompt:

Write me a highly optimized n-queens solver in go. It should take advantage of parallelism (what little there is) and output only the solution and how long it took.

After 10 minutes it gave me code that didn’t compile.

It took another 3 mins to fix the compile error and the output is not correct.

As I said - LLMs on 8Gig VRAM just aren’t worth it.

Yeah - I’ve been playing around more with the Qwen3-Coder-30B-A3B-Instruct MoE model and it’s still quite… Meh. I’ve been using llama.cpp and I’ve tried a bunch of tuning. It works and performs well enough (15t/s) but the output is just garbage. I can do some simple coding but I’m finding I’m fighting with it more than if I just wrote the code myself. Maybe I just have standards that are too high. Claude Opus 3.7 is just in an entirely different league…

I’ll check that out - speed isn’t my biggest issue so much as coding performance… The qwen 3.5 model I was using can write code, but it’s… Meh? Like sometimes it doesn’t even compile.

I did try tweaking llama.cpp to do some cpu offloading and it does seem to allow for much larger contexts at a modest performance loss. I’ll check out larger models.

64G. But CPU inference is painfully slow.

I’ve tried a few times but with only 8gig of vram it’s simply not worth it.

But there is a system and it sounds like it works. It just wasn’t in place. The problem is not a technical one.

Another factor in the crash was the fact that emergency vehicles at LaGuardia were not outfitted with a transponder as part of the airport’s surface surveillance system, known as ASDE-X. The system is designed to prevent runway collisions by creating a display air traffic controllers can use to track the movement of every plane and vehicle in real time.

That’s almost even more stupid. Does he know how quickly they reproduce venom? Does he know what dose he’ll receive? It might even be too little.

This guy sounds well meaning, but stupid.

Why let snakes bite him rather than extracting venom and injecting known quantities at lower (and safer) levels? That sounds like a remarkably stupid way to do this.