The Ghost in the Blog: Decoding False Memories

May 2025 already—and guess what surfaced? A digital time capsule from 2007, buried in old blog files like fossilized thoughts. Just yesterday, it was team dinners and product talk. Today, it’s ghosts from the flat-file past.

Blogging back then wasn’t just logging in and hitting publish. Platforms like LiveJournal and TypePad still roamed the web, with Blogger barely finding its stride. I used BlogEngine.NET—plain XML, scattered image folders. No databases, no dashboards. Primitive, but kind of beautiful. Everything was raw, visible, and weirdly honest. And it got me thinking—what’s really real? Files or blobs? Physical or digital? Maybe it’s all just perspective, all just ones and zeroes dressed in different metaphors?

But I digress. There was a problem. When I tried to open those old blog posts, the encoding was all off. The blog platform likely saved them using Windows-1252 or ISO-8859-1—formats that were standard around 2007, before UTF-8 became the dominant encoding—and now they read like a weird Esperanto remix. Letters were mangled. Accents went rogue—like a glitch in the Matrix where language momentarily forgot its own code.

Luckily, ChatGPT (yes, the trusty sidekick) saved the day. It somehow understood the mess, re-encoded the content properly, and even offered me markdown exports or Canvas drafts. I chose Canvas. Why? Because I wanted to see them side by side, feel the flow.

When AI Rewrites Your Memory

Most of those posts were travel logs from a West Coast U.S. trip, back when I was adventuring with a coworker in San Francisco. Sadly, I had long lost the photos and didn’t see the value back then. But I kept a few entries that still meant something to me.

One post, in particular, stopped me cold. Here’s where things get weird. GPT totally rewrote one of them: the McDonald’s story. Originally, it was about working in a computer store next to a McDonald’s. Somehow, the model turned it into a coming-of-age tale of me working in a McDonald’s. Totally different vibe. A different story. And not mine.

I caught it immediately. But it made me think—what if the AI had made only small changes? Ones I wouldn’t spot so easily? It turns out the issue stemmed from the token limit: I fed in all 12 blog posts at once, which exceeded the model’s capacity. As a result, it began hallucinating earlier entries—like the McDonald’s one—rewriting them based on what it could still remember or stitch together. Fascinating, but also a bit eerie.

And that thought collided perfectly with something I saw on LinkedIn this morning. Someone had done an experiment: feed the same AI image prompt over and over, with just one instruction—“Don’t change a single thing.” 70 iterations later, the image had morphed into something surreal, almost creepy. From realism to comic-book, from familiar to uncanny. At one point, it even changed ethnicity. Model collapse in action. The AI equivalent of faxing a fax of a fax until the message disappears.

But the real kicker? That prompt—don’t change a single thing—might be the issue. It’s like the old “don’t think of a pink elephant” trick. Just by saying it, you’ve summoned the change. AI, like us, struggles with negation.

So no, I don’t think AI should be used to create perfect copies. That’s a trap. A flattening. A ghosting of the original. But using AI to remix, to reinterpret, to pull in new data, and tell the story from a different perspective? That’s the magic. That’s the purpose. You don’t simulate truth; you simulate perspective.

Imagine an AI that combines the knowledge of a horticulturist, a space traveler, a marketing guru, and a sci-fi novelist—and then asks it to pitch a houseplant that thrives in zero gravity and glows with your brand colors on cue. That’s not creepy. That’s incredible.

Excavating the Archives

Speaking of incredible, let’s loop back to the blog archive.

The first post I rescued? Dated December 31, 2007. Reading it felt like opening a message in a bottle from my younger self—earnest, curious, and surprisingly familiar. The launch of Mike’s Mind. I wanted a cozy corner on the internet to share my thoughts. Back then, I was transitioning out of corporate life at Microsoft into something more independent. Writing better reports. Building better bulletproof presentations. Trailblazing stories. Capturing photos—though now, I’d say I was capturing philosophies.

The next day, January 1, 2008, I posted that McDonald’s story. And another one about spending New Year’s Eve alone. Not lonely—just alone. There’s a big difference. I remember it vividly.

Then came the “Digital Locksmiths” piece. Still blows my mind. SSL encryption keys? You have to pay for them. Yet when you buy a house, no one charges you extra for your front door keys. It was a post shaped by early frustrations in digital ownership. And it still rings true. There’s an opportunity hiding in plain sight there.

Then there’s a jump—to 2015. That’s when I coined a new syndrome: Jupiter Ascenderitis. Inspired by the movie Jupiter Ascending, it captured my creative crisis from trying to force too many ideas into too little time, like jamming a galaxy into a teacup.

Another post from 2016, titled “The Perfect Paradox of Paralysis by Imperfect Peripheries.” A mouthful, yes. But also a turning point. That’s when I shook off perfectionism. Pre-ChatGPT, but post-realization.

Funny enough, I looped back into my youth. Playing simulated 2D lives with Little Computer People, composing Amiga tracks, dreaming in digital bits. I had no idea who Philip K. Dick was back then, but I was already halfway into Do Androids Dream of Electric Sheep? territory. Pure coincidence—or maybe not.

I even mentioned that game during dinner last night. No one remembered it. That floored me. Little Computer People was huge! You’d load up the disk and a tiny character would start living inside your computer, eating, typing, pacing, even reacting to you as if it were a digital roommate with its own routine. It wasn’t even a game. It was a simulation. A slice of digital consciousness in a time before we had the words for that.

Drafts, Loops, and What Comes Next

To wrap it up, there’s more in the drafts—posts I teased and never finished. Like “Captivated Like Spielberg.” Or “Dan’s Law.” Stuff that’s nearly there. And now, with GPT’s help, maybe they’ll finally find their way out into the world.

So yeah, that’s today’s reflection. Like the ghost in my McDonald’s post, some stories return in disguise, reminding us that what we archive, we might one day reawaken—distorted or clearer than ever.

On blogs. On AI. On old stories, and the strange ways we retell them.