I get asked how I manage to keep this blog ticking over with tech stories, Home Lab tips, and personal experiences. The short, perhaps slightly uncomfortable answer is that I don’t do it all by myself. AI plays a part in my workflow. Now, before you roll your eyes and assume this is just another “dead” site filled with bot-speak, hear me out. For me, AI is a tool—a piece of software—not an invisible ghostwriter.
The final words, the actual technical logic, and the human perspective remain mine.
I’m a sysadmin, not a marketing lead. I’m also someone who fundamentally dislikes the “Big Tech” status quo; Microsoft and Google aren’t getting into my data if I can help it. Because of this, I’m extremely picky about how I use these LLMs. It isn’t about handing over my brain. It’s about speeding up the dull bits. Brainstorming, research, or getting a rough skeleton down when I’m staring at a blank screen are tasks for the machine. The final words, the actual technical logic, and the human perspective remain mine.
I can already hear the self-hosting purists: “Why don’t you just host your own LLM on your Proxmox cluster?” Look, I love my Home Lab, but I’m realistic. To run a model that is actually useful—one that doesn’t just hallucinate nonsense or take ten minutes to generate a single paragraph—you need hardware and a power budget I simply don’t have. I’d rather use my resources for actual VMs and Docker containers than trying to liquid-cool a GPU cluster just to write a blog post.
For this, I rely on Abacus.AI. I chose it because it actually respects privacy, which is rare. They don’t sell your data. More importantly, they don’t use your personal information, prompts, or uploads to train their own models or those of their partners. Everything is encrypted and processed for real-time inference without being stored. You can go and read their privacy policy if you don’t believe me; I did.
I also use AI for the visuals. As I’ve mentioned before, I’m a sysadmin, not a designer. My skills with GIMP (the FOSS alternative to Photoshop, obviously) are… well, they aren’t my strong point. I like my posts to be visually engaging. Without AI, I would have to spend hours hunting for Creative Commons images or navigating the legal minefield of permissions and licensing. It’s a massive time-sink that honestly doesn’t interest me. AI lets me generate a relevant image in seconds while avoiding copyright headaches entirely.
AI should be a power tool that complements a person’s expertise, not a cheap replacement for it.
It is 2026. AI is everywhere. It’s incredibly useful, but I have zero time for the “AI slop” currently flooding the web—pointless, low-effort videos and sites that are clearly 100% generated without a human synapse firing. I believe we are right to embrace this technology, but we have to use it properly. It should be a power tool that complements a person’s expertise, not a cheap replacement for it.
Here is how the workflow actually breaks down:
Idea generation: When I’m stuck, I use it as a digital whiteboard to brainstorm fresh angles.
Research assistance: Great for summarising complex concepts, though I always double-check the facts because hallucinations are still a thing.
Drafting outlines: I’ll occasionally have it generate a rough structure which I then flesh out with my own technical experience.
Image creation: It handles the aesthetic front so I don’t have to worry about copyright strikes.
I don’t let AI write my posts. It’s a helper. I’m still the one deciding what hits the “publish” button and what gets binned. Keeping this blog genuine and technically accurate is important to me, and frankly, a bot can’t do that.
Working smarter doesn’t mean compromising your values. It means using the right tool for the job while keeping the big corporations out of your business. If you’re looking at AI, treat it like a CLI tool: powerful if you know the flags, but useless if you don’t know what you’re trying to build.