Makes sense. Except I can’t really say my muscle memory suffers either. The workflow on my home pc is just so different that it actually occupies a completely separate section of my memory, rather than overwriting it.

Btw, since you mentioned you’re heavily relying on mouse in your workflow, i suggest you to take a look at Kando. It’s cross-platform, and looks super cool.

actually, it’s lomi-lominux

except it’s derived from Linus’s name, which in finnish is pronounced as Lee-noos, so technically it should be Lee-noox.

In reality tho? Nobody cares

I use KDE on my desktop because I don’t care about maximizing screen real estate in that context.

my whole screen real estate banter was in response to

I make XFCE as Unity-like as possible, which I believe is the best interface for limited screen real estate.

Any DE can become “hotkey-first.”

Yeah… no. Maybe i’m just too picky about my preferences, but ctrl+alt+t, alt+f4, and other “wonderful” hotkeys that require either a huge hand stretch, or two hands to press them, don’t sound like “keyboard-first” to me. Neither does setting up said hotkeys either with a gui utility, not really meant to be used with both hands on the keyboard, or editing an ugly auto-generated json or xml config.

Of course, all of this comes down to how one’s workflow is.

Of course, it’s nothing more than a matter of personal preference. I’m not trying to sell you anything, or prove your stance wrong. I just like talking about this stuff. Tell me if i’m being annoying.

I don’t ever see myself wanting to scroll my desktop or deal with windows auto-tiling.

The whole point of tiling/scrolling wms is that you have to deal with less things tho? The ideal goal, philosopher’s stone of those wms, if you may, is to make so that user, after the initial configuration, doesn’t need to concern themself with window positioning and resizing.

Multiple desktops is another example. It’s something just about every DE offers, but I’ve never had a use for it.

That’s because those DEs make it so that using them actually takes effort, because guess what, the default key bindings for them are horrible! On top of that, i’m not sure, but i don’t remember neither XFCE nor Plasma offering any sort of automation around it, (might be wrong tho) whereas in tiling/scrolling world custom rules for automatically assigning certain windows to certain workspaces, or, for instance, opening an app if you switch to a certain empty workspace, is considered a bare minimum. It naturally makes the process of navigating the windows even less chaotic. Basically, it creates a mental model, where you don’t have chaotically alt-tab around or seek anything with your eyes, since you just know where everything is. Your desktop becomes the perfect silent well-oiled machine that gets in your way as little as humanly (machinely?) possible, and it just feels soooo incredibly fucking nice.

I’m using the term “WM” loosely, as a more of a general term here.

And no, i can’t call a DE that adds a title bar to every window, has a huge-ass taskbar, and a limited amount of free space per workspace “best for limited screen real estate”. Something like Niri, MangoWC or Hyprland (with scrolling layout enabled) works much better and cleaner precisely in that regard, since they offer

a) a hotkey-first UX, meaning that everything usually represented by a UI button is moved to the keyboard and no longer needs to clutter the screen

b) literally infinite horizontal space.

I literally can’t go back ever since i gave tiling a try, any floating de/wm now feels clunky and awkward to me.

IMO, Windows’s UI/UX might seem good only if you haven’t tried anything actually different. Tiling/scrolling WMs is the way.

I tried doing this at first, but i just kept being frustrated every time something wasn’t like on windows…

Then i gave tiling managers (i3 specifically) a try…

Here I am now daily-driving Hyprland and writing custom ui with quickshell…

One out of ~12 windows versions not installing a driver = bad Only 3 Linux distros out of hundreds installing a driver = good

Windows server is not “one of ~12 windows versions”. It’s a windows distribution that exists in parallel with normal windows.

IIRC, microsoft added automatic driver installation only in windows 10, so even if i indulge your sloppy attempt at drawing equivalence between major windows releases and linux distros, it’s two for three, and not in windows’ favor. Except there’s no equivalence. Windows versions exist much like any major release of any OS distribution with fixed release model.

Imagine a distro so bad, it has no package manager (or any software version control, for the matter). It has no customization, supports only the most primitive and prehistoric file systems. It’s so bad that even the users that were conditioned to use it prefer to stay on a decade old releases of it because the newer ones are just that much worse. Actually, you don’t have to imagine, because that’s just windows. The only advantage windows has is it’s desktop market share, but even that’s slowly diminishing lately.

Also damn near every Linux distro has some sort of bloatware. There’s always apps I don’t want because I won’t use them or because I prefer something else.

Guess what? Unlike in Windows, you can delete all of it, just like literally any other program that you have installed. You don’t want it to be installed in the first place? Boy, do i have news for you, because Arch, minimal NixOS or even Gentoo exist just for you!

That’s the beauty of Linux, and FOSS in general. It gives you freedom. Windows only takes it, more and more. Each year windows gets slightly worse. Each year linux gets slightly better.

Ubuntu comes to mind as the biggest offender.

Preinstalled software suite is not even in top 10 of the things that are wrong with ubuntu. It’s just a bad distro. I’ll tell you more, despite whatever people are saying, Debian and all of its derivatives in general are quite less than ideal for your average desktop PC user.

well duh, of course it varies. Do you expect Windows Server to automatically install gpu drivers too? On CachyOS, EndeavorOS and Bazzite everything also works out of the box. And you don’t even need to disable one hundred and one microsoft-mandated bloatware when you install them.

uhm… you don’t have to hope…? You can literally check which driver is installed. You can even install preferable driver version via your package manager and delete mesa drivers if you don’t want your system to ever use the integrated gpu.

Lutris seems to be the solution for your problem. Also, for older games, often than not, you gotta try few different proton versions.

Except even this is no longer an issue. There’s nvidia-open driver, and, at least to my knowledge, it works without any problems.

What even is a “professional developer”? I mean, given that you aren’t an egg-headed RnD developing new technology for someone like Nvidia, the job of a developer is 95% mind-numbing routine and typical tasks solved by applying ready-made patterns.

i’ve heard of the dataset poisoning and degradation caused by llm-generated content present in the dataset myself, but i’m not sure whether it was a practical observation, or a mere experiment. And I still fail to see how new datasets are really useful for developing a new llms, or how it’s a problem for the devs to switch back to the older datasets.

And the cornerstone stays the same: to have any significant effect on the final LLM quality, shouldn’t the poisoned (either by llm-produced content, or by intentional poisoning) data portion be… well, statistically significant?

i think that sort of issues will no longer be relevant, considering that there are official semi-opensource nvidia-open drivers and that even wayland supports nvidia gpus just fine nowadays.

except for poison to take in, it should be a pretty significant part of the dataset. Also, ngl, i’m not much informed on the topic, but aren’t all the datasets, if we’re talking about generic diffusion models and LLMs, already been formed? From what i gather, the innovation in AI mainly comes from utilizing new architectures, rather than training a model on something unique.

not really related to the post, but NGL, it kinda fascinates me, how with the advent of LLMs, they became the ultimate punching bag for whenever something works bad, as if people didn’t write even more horrible things without them (my regards to javascript and python).

just to clarify, he’d not so much called LLMs “the new compilers” as he compared both to each other in a sense that an LLM is just another layer of analysis tooling between the developer and the final machine code, which, IMO, sounds much more reasonable than calling LLMs “the new compiler”.

it’s not so much “bending” as it is treating ai as a potentially-useful tool when done right. Which it totally is.