I've finally done it. After almost 30 years of Windows I have finally taken the full plunge into Linux. And you know what? I should have done it much sooner.Let's back up, what's the problem with Windows? Windows is but another victim of modern enshitification. Once a product gets popular, it becomes more appealing for companies to try to milk the product for all its worth and to continue riding the momentum of success. As with Windows, its been getting more and more intrusive, filled with ads and various crapware, which lead me to think more and more of the greener pastures. Sure, for a while I continued coping and make concessions such as by switching over to Windows 11 IOT LTSC, but ultimately it's been feeling like I've been climbing up a sinking ship. As I now have a bit of free time I've decided that it's finally time to take the plunge and to jump fully into the world of Linux.So which distro did I choose? As historically an Ubuntu (Kubuntu is better) user, I wanted to try and find something better. My research lead me to Linux Mint, which is arguably the most popular distro right now due to its maturity and reputation of using old but well tested software for the sake of stability. Its desktop environment, Cinnamon, is highly praised, but ultimately it did not work well for me and its lack of Wayland (a rabbit hole not worth jumping into for a casual) support and slow release cycle were too frustrating. I had inadvertently stumbled into a Firefox bug where it would not match the system theme unless the theme was toggled a few times. The bug was identified and fixed, but I'd need to wait for the next bi-annual release of Mint to receive it. After a bit more research I went with the fad and installed a *drum roll*...Arch Linux! Not to sound like too much of a glazer, but Arch + KDE was easily the most enjoyable desktop experience that I've ever had. While KDE Plasma 6 has full Wayland support and endless customization, should one choose to look for it, Arch's simplicity and flexibility was the key for me. Consider that while Mint officially supports 3 different desktop environments, Arch supports around 14 and the user does not need to re-install the entire system. Its package manager, Pacman, is a joy to use and its AUR (Arch User Repository) provides access to a vast array of managed software that encourages users to stay closed to the stable but bleeding edge. Arch allows for full flexibility and even allowing the user to select the kernel (I went with the default), which is currently about 7 minor releases ahead of Mint's. The Arch Wiki is one of the top resources for Linux users right now and its community, system performance and momentum are currently unmatched even by the likes of Mint. I'm not even (yet) going to dip my toes into the world of emerging tiling window managers, like hyprland, and desktop "ricing", but never say never.While perfect software does not exist, what Arch and its community is creating something truly special. I so far very happy with the decision to switch over to Arch and I hope that it continues to impress me in the future.

The year is 2015, and the top summer blockbuster is a sci-fi film featuring a man-made AI named Ultron that tries to take over the world and kill all humans. Fast forward to 2025, and AI is back with a vengeance, but this time it’s not a fictional character. It’s a reality that has transformed the way we live, work, and interact with technology. In particular, the rise of Large Language Models (LLMs) has ushered in a new era of artificial intelligence.To me, there are many pros and cons to the takeover of our robot overlords. The pros are largely focused on the rise in productivity in areas such as software development, where LLMs are being heavily utilized. In fact, I'm having a hard time writing this blog post without peeking at what the LLM is trying to guess I’m attempting to write next. At times, it's humbling to read that an LLM can write something so much more eloquently than I can. It’s dizzying reading what it thinks I should write next but instead fighting to express myself as I want. It’s a little eerie—which brings me to the cons. Outside of the tremendous energy consumption required for training and running LLMs, they are dulling our ability to think and perform for ourselves. The long-term consequences of this could be catastrophic, as the value of human creativity and mastery is on the decline. Why work hard or try when an LLM can generate it for you? What will happen when the world is filled with AI slop that will be used to train the next generation of LLMs, which will create even sloppier content? I can’t even begin to guess, but my goal is to continue using LLMs as a tool to improve my life. I understand that I must be highly skeptical of anything it generates for me and especially skeptical of generated content in areas where I am not an expert.As we navigate this AI landscape, I encourage one to consider the implications for privacy. In today’s world, corporations chase the "big data" dollar by collecting as much user data as possible to build profiles that can later be sold to advertisers or monetized in other ways. My journey toward privacy has led me to , where you may find all sorts of open-source LLMs that can be run locally on your own hardware. Some lessons to be learned:

• The environment in which you run your LLM matters. For me, it was the ease-of-use and general UX that made the process much more enjoyable. I initially started with koboldcpp, then Ollama with Open WebUI, then I tried LM Studio and finally I came all the way back around to koboldcpp. While koboldcpp may not have the best UI, it is the most performant and flexible based on my testing on both Windows and Linux.
• LLMs will run fastest on your GPU rather than your CPU, and your video card brand matters. Right now, NVIDIA is the go-to for most LLM tasks, thanks to its proprietary CUDA libraries for GPU acceleration. While CUDA is the leader, it’s not the only option. AMD has been making strides in the AI space with its competing ROCm platform. Alternatively, you may also run Vulkan or CPU-based runtimes, but they are just not as performant.
• The size of your GPU's VRAM matters a lot. This is because for an LLM model to run at peak speed, it must be fully loaded into memory. For example, a 70B parameter model will require significantly more VRAM than a 7B parameter model. If you have a high-end GPU with 24GB of VRAM, you can run larger models like Llama 3 or GPT-4. If you have less VRAM, you may need to use smaller models or quantized versions. How I wish I could go back in time and buy a GPU with more VRAM!

As I continue to explore the world of AI, I’m excited to see how it will evolve and improve. I would love to be able to run agentic models locally, but I have yet to find a way that may rival the performance of CoPilot, Cursor, etc...

I screwed up. I really, really did. I took my privacy for granted and now I'm feeling legitimately creeped out that a single monopolistic, advertisement conglomerate knows basically everything about me. Where I go, what I read, what I buy and basically most of the significant events in my life. Now don't get me wrong. I have nothing to hide, but I also don't have anything to share (without my express consent of course). Up until last year I had an Android phone in my pocket, browsed the web using Google Chrome, sent all my emails using Gmail, noted my events in Google calendar, had Google Analytics on all my personal sites, uploaded all my pics and videos to Google Photos and stored some personal files in Google Drive. Notice how many times Google comes up there? It's absolutely insane! Google services are just far too convenient and most importantly free, which make them very attractive to use for many people, me included.I am aware that the world is going through a privacy crisis where the 14 eyes have made closed door agreements to record every moment of our lives in the name of "fighting terror" or for "national security", but the reality is it's all about amassing power. I don't need to write too much about this as Edward Snowden put it all out there for the world to see, but most still aren't listening. While I do think that if many entities and/or corporations knew bits and pieces of a person's life, that would not be so scary. It becomes very scary when all the data about an individual gets aggregated together given that all western countries have virtually no privacy laws (or at least none that get enforced) and the state is above all.I am now fighting to take my privacy back by:

• Only using Firefox as my personal browser. In Firefox I have enabled enhanced fingerprint prevention (hidden feature meant for Tor users), enhanced tracking prevention, total cookie protection, HTTPs everywhere and I use uBlock Origin in "easy mode" (to block third party scripts and frames by default).
• Using an iPhone and never Android. Maybe in the future I will get an Android device and flash a more privacy-focused OS onto it.
• Absolute no Facebook services or any kind of social media. However, for the time being, I am making an exception for Strava which is social media adjacent.
• Switching over to Protonmail for my email, calendar and file storage.
• Removing Google-analytics from all my personal sites.
• Switching to use startpage.com for searching instead of using Google directly.

I encourage the reader to think about their privacy and how comfortable they are allowing mega-corporation and the state knowing and recording all aspects of their lives. And always remember, if a service is free then you are the product!

It's been some time since my last blog post as I was busy moving and settling into my new home. It's given me the opportunity to discover the magic of home automation and right now the name of the game is Home Assistant. Through Home Assistant, one may run a local server (I run it on a Raspberry Pi 4) that may connect to smart devices and unify them all in a single UI. The greatest benefit of this is that now these devices may be automated together and they may now be controlled remotely even when away from home. Another amazing thing about Home Assistant is that it allows the community to extend its functionality through the Home Assistant Community Store (HACS). HACS essentially allows the user to download custom device integrations and UI customizations from GitHub. Sounds cool? Because it is! My current automations are as follows:

• Turn off all the lights, change the alarm system to "Away" and start the Roomba when all home occupants have left
• Re-dock the Roomba when one of the home occupants comes back
• Change the alarm system to "Night" after 10pm
• Change the alarm system to "Home" after 7am unless no one is home, or when someone comes back home from 7am to 10pm
• Send an alert to all home owners when any of the wetness detection sensors turn from "Dry" to "Wet"
• Flash all lights red when when the router disconnects from WAN for over 3 seconds
• Turn the lights on to 40% in the living room when a home owner comes home after sunset

Now that the summer's here, I'll be looking to automate my A/C using an IR blaster and to automate the positioning of my window blinds. More about my dive into home automation in part 2!

My faithful Pixel 3 endured 3 years of heavy use, but it's finally come time to put it to rest and so came the time to pick out a new daily driver. In my opinion, phones should be compact and I am absolutely not a fan of the fact that all the modern Android flagships are at least 6.2" in screen size. Year over year there is a very real push to create faster, higher spec phones and the simplest way to do that is to simply create larger phones so that they can hold higher performing components. As such, this time I chose to go with my very first iPhone. Mind you, this is not my first Apple device. In particular, I'm a big fan of AirPods and I actually rocked the very first iTouch back in the day. This time I went with a (product)red iPhone mini and it's been fun getting used to a new phone operating system. I consider myself an advanced user and I've now spent a ton of time tinkering and playing around with my new phone. Here is my comparison of the pros and cons of Android vs iOS.iOS Pros:

• Beautiful screen and a powerful haptic engine
• The UI generally looks better
• Significantly less bloatware and most apps that come pre-installed can be uninstalled
• AirPods work much better with their intended OS. iOS works much better with the Apple ecosystem devices
• Widgets for the stock apps look great
• Iconic phone design and sounds

iOS Cons:

• It really hurts to lose Pixel's call screening
• No system-wide back gesture results in an inconsistent experience between apps
• Terrible UX without an iWatch due to the new 6 digit passcode, lack of touch id and severe limitations of face ID for this COVID season
• No USB-C :(
• Live wallpapers are implemented in the worst way (why do I have to press down on the screen to animate the wallpaper? why???)
• Much harder to block system-wide ads
• severe custom app restrictions (I ended up paying $30 a year for a appdb PRO subscription which allowed installation of "non-app store" apps)
• The stock keyboard is horrendous and does not allow much customization (custom keyboards come with restrictions which make them not ideal)

There are definitely a fair bit of shortcomings of iOS over my previous Pixel phone and it was not easy to get used to at first. Since then I have built up my muscle memory for re-entering the passcode and it's starting to feel better now. I've decided I'll be be sticking with my new iPhone mini for the foreseeable future.