Looking in that thread it seems to be largely small scale/fine tuning where the data is cherrypicked… they’re not considering the corporations which create the models they’re finetuning in the first place

That’s why we invented bridges and viaducts, we didn’t want the trains to feel left out

Exactly, 60k rows is negligible enough in most cases that you can just treat it as free unless you’re doing a cross join on it or something, unless he’s doing something like using an unordered text file as his database with no ram or cache

This is Rust not Python

They’re both optimised out by the compiler. If you disable compiler optimisations, they’re identical in machine code anyway, unless you introduce a second loop, in which case the first will be more memory efficient as the memory used in the first loop can be reused in the second loop, whereas if you declare the variable outside the loop it can’t (again, without compiler optimisations, which make the whole comparison pointless).

if they mattered they’d be errors I’m sure

C is memory safe if you program it well enough, so I guess C

Yeah even gpt4o couldn’t keep track of encounters, run battles etc. in my case…

I think if you wanted to do it mechanically consistently you’d probably need to integrate it into a vtt where you give it context and potentially fine-tune it to give quest related summaries & gming rather than just “stuff”

Yeah, of course it varies place to place but I think for the majority of at least somewhat developed countries and urban areas in less developed countries 50Mbps is a reasonable figure for “normal home internet” - even at 25Mbps you’re looking at 4½ hours for 50GB which is very doable if you leave it going while you’re at work or just in the background over the course of an evening

Edit: I was curious and looked it up. Global average download is around 50-60Mbps and upload is 10-12Mbps.

LLMs have a very predictable and consistent approach to grammar, punctuation, style and general cadence which is easily identifiable when compared to human written content. It’s kind of a watermark but it’s one the creators are aware of and are seeking to remove. That means if you want to use LLMs as a writing aid of any sort and want it to read somewhat naturally, you’ll have to either get it to generate bullet points and expand on them yourself, or get it to generate the content then rewrite it word for word in a style you’d write it in.

Intel Arc also works surprisingly fine and consistently for ML if you use llama.cpp for LLMs or Automatic for stable diffusion, it’s definitely much closer to Nvidia in terms of usability than it is to AMD

Docker fan mindset

What OS doesn’t do that, even linux has xdg dirs

I just find the saving mechanism frustrating to use compared to vim’s as an entry level user, and now as a mid-skilled user I dislike how featureless nano is - when I was first learning how to use the terminal I hated having to edit anything as I was pretty much force-fed nano with no alternative provided, but on finding vim and remembering literally 3 things (:w, :q and i) everything became so much easier, but I definitely do have an extra bitter taste left about not being told about something much easier to use which irked me when I saw someone preaching how amazing nano is

I also really don’t get the hate for vim when remembering 3 things gives you as much/more functionality as nano and is a starting point for so much more functionality - intuitive doesn’t mean featureless and don’t try and pretend nano’s shortcuts are the same as 99% of other editors (text or otherwise), in fact they’re totally different, making it less intuitive

Nano is just as fiddly as vim and way less powerful when you actually figure out what you’re doing though?

Ie a completely redundant piece of software that has no place being pre-installed anywhere

It’s not so much a circlejerk as much as a knowledge that KDE plasma is the most approachable DE with the most polished first experience for the majority of new users

The reason it gets interpreted as Gnome bad is that both Plasma and Gnome both mainly target users who want something that just works out of the box and doesn’t have a steep learning curve, however KDE have managed to keep up better with what new users want in recent years while Gnome has fallen into a semi-trap of doing what their current/older users want. That doesn’t mean it’s a bad distro, frankly it’s great for their current users, however it does little for newer users who may not find it as intuitive as other DEs, therefore making it a worse default DE for “off-the-shelf” distros targeting new Linux users.

At the end of the day though, it is about personal choice, and nobody’s saying i3 isn’t better for powerusers or that LXDE doesn’t run faster, but if you have the knowledge that you want to install one of those or the many other DEs available, then you can just find the iso/distro/package with that DE and install it rather than just clicking the all-in-one-guaranteed-to-work-lts download button on the distro’s homepage

The syntax is certainly easier than Java

And VisualBasic’s syntax is easier than COBOL, but this isn’t a competition to make the least offensive heap of putrid garbage, so why does it matter?

Python works just fine for basic scripts, frankly it’s amazing for it, but oop and functional programming is so incredibly obviously badly shoehorned in that huge swathes needs scrapping and version 4 releasing

I mean once you get beyond bash-like scripts python is esoteric as fuck, adding oop to what is essentially a shell is a terrible idea

That said, there’s plenty of languages with good syntax that is still good when you get into more complex stuff (modern C#, scala, kotlin and more)

else { SneakilySecretlyCollectData(user) }

Yeah, gleam.io really fucked up I guess…

Unironically though people asking questions, then further explanation, then posting when they figure it out is pretty optimal compared to above average documentation