This is Part 1 of a multi-part article

I’ve been a long‑time AI skeptic. I’ve used it casually over the years, but was—let’s say—less than impressed. Early AI “assistants” never felt useful in any real‑world context. Platforms like Claude and GPT had potential, but they were inconsistent, rough around the edges, and getting exactly what you wanted felt more like luck than design.

And honestly, the hype made it worse.

Tech companies forcing AI into everything (looking at you, Microsoft), questionable data scraping, privacy concerns, environmental and economic impacts (have you seen RAM prices lately?)—none of it helped the case. It all felt like a tech bubble inflating itself on vibes and marketing.

From where I stood, AI looked overrated, overhyped, and underperforming. A bubble destined to burst.

“Fine… I’ll Try It Seriously For Once.”

Eventually, I hit a point where I needed something done quickly and didn’t feel like doing it myself, so I gave AI a real shot — not a casual test, but an actual attempt to use it for something practical. I started with the low‑effort stuff: summaries, quick explanations, small research tasks. Things I normally procrastinate on.

To my surprise, the newer models handled these tasks better than I expected. They were faster, clearer, and noticeably more reliable than the older versions I’d tried. Nothing mind‑blowing, but definitely competent.

And competence goes a long way when you’re staring at a pile of things you don’t want to waste brainpower on. I still wasn’t convinced AI was “the future,” but I found myself opening it more often — not because I trusted it, but because it was simply the easiest option.

It wasn’t impressive enough to convert me, but it was enough to keep me curious.

“How Do I Use This Better?”

As I used AI more often, I started asking myself, “Am I doing this right?”

Were the mediocre outputs really the AI’s fault — or was I feeding it mediocre inputs?

AI is “conversational,” sure, but it’s still a computer. Wording matters. Context matters. Structure matters.

So I did some research.

Turns out: yes, prompting matters a lot.

I took an intro‑to‑AI course and learned some better prompting techniques. The improvement was immediate. Not perfect, but noticeably better. Once I realized AI could handle the small stuff, I started pushing it a little further — explanations, breakdowns, concept reviews, documentation shortcuts.

It became genuinely useful in my daily life, not just a novelty.

This was the moment where I shifted from “skeptical user” to “curious experimenter.”

“Well, What Else Can It Do?”

As my prompting improved, the obvious question hit me:

What else can this thing do?

So I started experimenting with more advanced models, image generation, and other AI tools. Some of it was a bust. Some of it showed real potential. And some of it surprised me.

One area that stood out was log analysis.

Drop in a giant log file → get a clear answer.

Still needed sanity checks, but it was definitely faster than reading 5,000 lines manually.

Then I decided to try its coding features.

I’m not a coder or a dev. I know a little — enough to hack my way through some program mods, make a couple small tools, and do the occasional “programming 100‑level” stuff — but I’d never written something from scratch.

And that’s when I found myself doing what the cool kids apparently call “vibe coding”.

Stay tuned, that’s where Part 2 begins.