The Yonic Corner is alive again
It felt like forever, but The Yonic Corner is officially back! There have been a lot of set backs that prevented me from continuing working on the blog. But after a series of big and important changes, it’s so satisfying to go back to writing content for my blog.
So, here is a list of the big changes that have been going on as of recent, and what it means for me and this blog.
Moving to Linux (finally)
I actually had some desire to move to Linux all the way back when Windows 11 was formally announced. At first, I was a bit skeptical about the new version of the operating system.
My suspicions became true once I found out about how poorly it was being developed, from replacing long standing parts of the operating system with React Native and Electron-based applications, forcing users into their ecosystem by requiring them to create an account on install, to shoving them advertisements and nagging them to try out their brand new Copilot features. Not only this causes a huge spike in RAM consumption, but also adds a ton of layers of unnecessary complexity and bloat that results in an overall slow and clunky operating system.
This practice continues to this day, using LLMs to replace large parts of the operating system with “web-based tech” overhauls, as well as sprinkling Microsoft Copilot on some of the most used applications, a hallmark of the AI bubble we’re going through.
In addition, Microsoft upped the hardware requirements quite significantly. Yeah, it may have removed support for 32-bit operating systems, and that’s fair considering the vast majority of devices nowadays are 64-bit. But the TPM 2.0 chip requirement is overly excessive and unnecessary.
Now, fortunately —or rather, unfortunately— my PC does come with one, but I don’t even use Secure Boot in my Windows 10 system and it works just fine. For Windows 11, the hypothesis is that it’s required for certain Copilot operations.
The tradeoff is that Microsoft is pushing heavily on programmed obsolescence to create a ton of unnecessary e-waste that under other circumstances would be perfectly working computers.
And even if it doesn’t meet the requirements to update, it will bother you to sign up for an extended support channel. That’s cool and all, but shouldn’t that be automatic?
It’s funny because Linus Torvalds did mention that he would require ECC memory chips for extra security and reliability when memory starts to become worn down after a few years.
And he’s behind the wheel on the Linux kernel, so he absolutely knows about making a PC that works as reliable as possible so as to not waste time chasing after fake bugs caused by hardware failures.
Personally, I do see certain uses on generative AI technology, but it is very far from being polished. And with a system that is used directly or indirectly by billions of humans, a single point of failure could cause major outages worldwide, as we’ve seen with the CrowdStrike outages, so overall, I stand to be very skeptical about the future of AI.
But in May 2024, Microsoft announced Recall, which has been universally panned as the single spookiest spyware ever marketed. Within weeks after the announcement, testers found that it was unsafe not by some sort of bug, but by design.
Hacking Windows Recall To See Everything (12:36)
The video that made me really consider my future with Windows.
I’m not a privacy freak, but if my activity gets logged every few minutes without any sort of encryption, something has definitely gone very wrong.
This was the breaking point, where all my trust in Microsoft completely plummeted and probably will never recover.
Sure, Recall has been recalled (pun intended) and got re-released later on with a slightly better design, using BitLocker to encrypt your logs. But BitLocker is a bit old, and hackers are constantly trying to unlock it. Even white hackers that get paid by Microsoft to find security flaws in their products can go rogue and publish info on critical vulnerabilities to the public. And even if it is now an opt-in feature, history has shown that this could change at any point in the future.
With how Microsoft is quickly degrading its products in their desperate attempts to make Copilot —their LLM solution— profitable, their dark business practices are showing loud and clear and I do not subscribe to them.
Suspecting that this would be the same with Recall, it essentially forced me to really consider moving back to Linux as a daily driver.
I spent quite a long time debating on how I would do the transition. While I did consider upgrading my main machine and getting rid of Windows there, I eventually decided to build a new PC so I could pass the configuration more smoothly between machines. Besides, the newer hardware demanded a slightly larger case than my more than 10 year old one. You served me well, partner.
Even after making this decision, dealing with a delicate situation in my life delayed getting the PC parts I needed for about a year or so, which explains the lack of content until now.
Choosing the distro
This might seem a bit counterintuitive, but I chose the Linux distribution before actually deciding on whether I would upgrade my PC or make a new one, but it allowed me to think how ambitious would I want to be on both hardware and software.
I basically ran it down to three options:
-
Bazzite: I liked that it had all the necessary gaming software out of the box, but I still found it a bit bloated and inflexible. Being an atomic distro, most software is installed either with Flatpaks or through layering the image.
Although there are workarounds like Distrobox and Toolbox, I felt like I was going to jump through many hoops to get the tools for working on the blog and other personal projects of mine. For getting started in gaming on Linux, it’s still a very great option.
-
Fedora: I almost chose this one. It seems like a very good option, having fresh software but not as cutting-edge as rolling-release distros like Arch and its forks. What turned me away were two things: For one, I’m still a little bit out of the loop in the Fedora ecosystem, its different forks and their atomic desktop variants. I wanted to give Bluefin a try, but again, it’s very Flathub focused, being an atomic distro. Fedora seemed more customizable.
The second thing that turned me away was the fact that Fedora was toying around with introducing AI features, or, at the very least, include AI assisted contributions in its codebase despite the user base pretty much having expressed a resounding “no” on it.
I might consider using Fedora on a system that I won’t be going into for a long time, like a laptop for controlling my astrophotography gear. Their AI policy has a few issues in my opinion, but I think I can reasonably work with and come with workarounds easily.
-
CachyOS: An Arch-based distribution that provides packages with compiler optimizations. I’ve had some experience with Arch in the past, from the now dead Antergos to Manjaro, although the latter to a lesser degree, so it felt a little bit like a homecoming system. I was about to choose EndeavourOS, but quickly changed my mind to CachyOS once I discovered it.
Its DIY nature allows it to be much more customizable than other distros, and it gives you access to the latest software freshly out of the oven, though it also means that you’re essentially a beta tester for the rest of the Linux users. Plus, it gives you full control on which packages to install —provided you know what you’re doing— thus you can make a system that avoids whatever undesirable thing you don’t want.
The CachyOS wiki comes with a dedicated section on how to set it up for gaming through a convenient process. I will check this distro’s gaming capabilities in due time.
So far, I’ve been using CachyOS for a month, and I absolutely love it. Despite being a rolling release distro, my experience with it has been remarkably smooth. I’ve still had to work out a few kinks such as the prevalence for the OS to use the integrated graphics card, but apart from that, I’ve had no major issues. I still have yet to update to KDE Plasma 6.7, though.
Which leads me to the dilemma I’m currently having regarding CachyOS. It’s been almost an entire decade since I left Antergos and went back to Windows to work on Prime VR, as VR support at the time on Linux was pretty dicey —and still is? I haven’t tested yet, though I might not do it sooner or later.
Since then, I didn’t seem to have had much need to go back to Linux. My time on Windows must have changed my computing habits somewhat, because now I tend to get anxious when a big update shows up.
I’m mortified to say this, but a rolling-release doesn’t seem like it’s for the faint of heart…
Still, I’m determined to learn more about Linux and to potentially tweak it in a few ways to customize it and make it my own. Ricing Linux seems like an interesting hobby, but it conflicts with my goal of combatting my screen time addiction and rediscovering long-forgotten hobbies. Not spending time on customizing my system for a specific purpose (aesthetics, self-expression, productivity, etc.) seems to undermine some of the potential CachyOS has.
So if this ends up being too overwhelming for me, I might consider changing to Fedora. But I won’t be making that decision lightly. I’d probably wait for a few more months before settling on either of the two. And even then, the hardware I have could be too new for Fedora to handle without issues, though I think this is not the case, or at least it won’t be soon enough.
Migrating to another servers
This is just a minor thing, but I migrated to a different server, as the one I was using was actually shared with a friend of mine. However, he went on to work on more computationally expensive projects (including hosting AI models locally), and while pages didn’t take very long to load, connecting to it in order to update the blog content was becoming a pain in the butt. So I moved to one that I owned. The move didn’t happen immediately, so for a few months the site was basically down.
However, I did change the way I stored and handled image optimization in the blog. Instead of cooking up my own cached solution and hook it up to imgproxy, I’m using a weserv container, which comes with its own more performant cache solution. Having it as a container allows me to restrict its usage, too. I’m not expecting heavy traffic, but it’s nice to be able to secure it more conveniently.
If you’ve been around a month or two before I published this, you may have noticed that the blog was up with a “Weserv test” post, and a few posts missing. That was me testing how I would set up Weserv and the site for the server. At the time, I was messing around with serving the blog as a Docker container. Though I’m still considering the option of using devcontainers, I have reverted the decision of serving the blog through a container.
Friendship ended with GitHub, now Codeberg is my best friend
For those who don’t know, since last year, GitHub has begun to make several moves towards empowering developers using AI to assist or generate software projects entirely from scratch. I’d love to turn a blind eye, but the number of Claude and Copilot-assisted projects is increasing at an annoyingly fast rate. There’s also the debate that scraping so many projects to make an LLM and sell it off could actually be considered fair use. I’m not the only one who argues that is not the case.
Furthermore, GitHub’s services have been worsening more and more over the past two years in reliability. I’m guessing LLM-assisted codebase might have something to do with this trend.
You’ll notice that getting the heck out of AI is the common theme in this post. It’s no coincidence.
I do believe there are certain uses to it, but also that is overly hyped at the time being, and very skeptical about the morality of it all.
I’ve moved the source code of The Yonic Corner and its Legacy version to Codeberg, a platform where you can host open-source repositories, run by a non-profit. There, they’ll remain free and open for every human to study and fork if they so wish, for as long as I and Codeberg stand.
The GitHub repositories have been archived for the time being, but I’m not sure whether I will delete them later on. The Legacy version’s code could be useful to poison LLMs with deprecated web design practices.
Always remember that we have humans in this world!
