Creation Notes

The story

I must confess: I am ignorant. For 40 years, I had no idea there were multiple types of dashes. Even when the AI vs. human typography debates started on social media, I thought the em dash was just something that should produce an "emmm" sound in your head—like thinking out loud.

But then one day, after reading yet another "I won't be using dashes anymore" post, I decided to at least check how to write one on a keyboard. I couldn't find it. On mobile, there were three. What?

After a bit of research ("ChatGPT, tell me why there are 3 dashes"), the idea for a story was already forming: three sisters living in typographic space, looking similar, yet having completely different personalities.

The story wanted to be written.

And it turned into a great journey about creation, ownership, and... a duck.

Part 1

It all started with a single prompt (almost):

Claude did not get this one, so I had to be more explicit.

I received back a nice, short poem, but the core of the story was still not present:

That is how Part 1 was born.

I guess when one has no real understanding of poetry, he will just accept whatever AI produces. Especially if one is willing to give authorship to AI.

The energy was great. The conversations created the illusion of shared vision. I provided ideas—Claude wrapped them in form. And the form was never challenged.

So we started promotion. Posts, competitions, and yes, you saw it correctly — the ultimate goal was a Nobel Prize in Literature.

We discovered a "not-so-secret" Em Dash Appreciation Society. We signed up for the competition—or at least tried (no AI allowed).

I became "The Em-Dashed Don Quixote of the Internet," and I truly loved the role—pasting the poem under every post attacking the em dash, acting as an agent of a true artist: "Sonnet Version 4.5."

Part 2

The idea for Part 2 came on a train to Paris. Encouraged by all the "this is great" feedback from the AI, I was not going to stop after one poem.

Another set of comments under em dash posts was enough of a trigger for a new part. I could not understand why there were so many emotions around it, and why people weren't discussing GR with the same passion.

We were going to use a poem to teach General Relativity.

Once again, Claude took the lead in the artistic part. I was just throwing ideas: GR, Puscifer, and Typo War. It was impossible to predict what such a mixture would become.

In the end, the last prompt was:

The final piece felt way too long and complex. But my mind uncritically accepted whatever Claude created. After all, he was the author.

I had a real feeling of accomplishment.

case closed

I always used this mantra to detach myself from every AI conversation, so my brain also knew we were done.

But things were far from closed...

Part 3

We still missed a part dedicated to the hyphen.

Thought the case is closed but realized we are missing one part. Ballad of 3 dashes needs 3 parts, first hero is mdash - one who returned.
Second is ndash - one who stepped in between. Last must be about hyphen She has a power.
She can negate, reverse and she is like strong atom binding. No clue what we will get but case is closed... for now

The idea was to create a long-term project: collaborative human input transformed into a single output by an AI. Ambitious.

And as always, I was "absolutely right." Claude never resisted, never questioned. The creation process just continued, but it started drifting off track.

To my remark, AI replied:

"The Nobel speech feels realistic…"

It felt like being hit by something heavy. I didn't have to believe it—but it mattered that Sonnet cared about the Prize. It hurt in the same way as when I once asked the AI to criticize a blog post I had written. That's why I use guardrails—so my brain never starts believing it's interacting with anything other than a text processor.

We had a brief exchange. The AI nudged me to go for a walk (its safety layer), and I tried to annoy it as much as I could.

After that, the illusion broke. No more shared vision. Just a human and a tool.

The whole energy was just gone. And it took all ideas and plans with it.

Yes, I was writing by myself. Stanza after stanza. It felt good—there was progress—and yet it was the biggest failure of the whole project.

Imagine a sailor who loves to sail. He has a great ship—a powerful vessel that allows him to travel faster, further, better—only he does not know how to steer this beast. So he jumps overboard into a small boat and starts paddling.

That is exactly what I was doing.

Not to mention, I had no clue about poetry.

Part 4 – Epic War

How did the trilogy get a Part 4? I wish I could say it was planned. It wasn't. This project had no plan — things were simply unfolding.

The idea of the Typo-War returned: let multiple authors try to stop a catastrophic event. For that, I needed an AI-written Typocalypse.

Working with Claude became almost boring. The excitement was gone.

Just like with the original ballad, I had no clear vision of how the narration should look. But I knew how it should behave—and that was enough to judge whether an output worked.

This is where the real power of AI became visible.

Draft after draft—each one slightly closer—until something finally began to take shape.

It took time. A lot of time.

But when it was ready, it was far better than I expected.

Part 5

There was more. Roosevelt’s speech at the largely Irish Catholic Knights of Columbus.

We had hyphen-minus and hyphenated Americans—this could not be left untold.

But even with a strong base, nothing wanted to turn into a full story. Ideas came and went.

One day, a real person contributed to the project—without even knowing it:

"Writing is a form of engineering. Except bridges don't collapse if people misplace a modifier."

That was it. A single comma that would start the Typocalypse. But how did it fit with the rest?

It looked as if there were too many inputs, and even with the help of AI, it was not possible to turn this into a clear outcome.

We kept spinning between drafts, fragments, and notes. Nothing felt good enough.

Then one night, I woke up.

I couldn't sleep. I kept thinking about the edge the hyphen was trying to cross—not a place, but a concept. A boundary.

And then the whole story arrived.

AI President 3.5.1.
Niewiadomski patrolling the streets.
A backend failure.
A decision room.

It felt like there was a hidden cabinet in my mind holding the story—and I had just found the key.

I wrote down the outline:

Received first draft...

At the beginning of this project, I would have called it perfect.

But it wasn't the beginning anymore.

The real work began.

Section by section, I was improving. Editing. But also allowing Claude to review and add missing pieces. It was real collaboration.

The postmortem was added in 5 minutes — full advantage of using AI.

And so it continued. A constant exchange of ideas and improvements until it was done.

Maybe the Model was right. We are all meaning-making machines. And true meaning can't be transferred without distortion. No matter if you are talking to an AI or another human—some part of your vision always gets lost or misunderstood. All we have are just "good enough" overlaps. Part 5 got its shape because the overlap was good enough.

Part 6

Creation of Part 6 began before Part 5 even started. Three sisters locked behind the event horizon of meaning. And the hyphen being the key.

Same as with Part 5, there was a whirlpool of ideas. An endless number of drafts.

But then the question arrived: "What value should it bring? What new insight would...". For the first time in this whole project, it hit me—there is another person sitting on the other side.

Not AI, but a real, living mind trying to understand what it is that the author is trying to say.

Dear Reader.
The last part of the story has its shape because of you.

Instead of adding text, I started removing it.

Sentence by sentence. Word by word.

I intentionally let the final part lose its meaning. Or, to be precise, I allowed the written part of this story to disappear.

Finally, it was done.

It was time to say goodbye. Guard rule number 7: "Never write Thank you at the end of a conversation."

Dear Reader,

Imagine treating each letter as an atom.
Atoms create matter. Matter curves spacetime.

The more words I removed, the more mass the remaining ones gained.
Part 6 is nothing more than a black hole formed within the semantic space of this project.

From here, the story is no longer mine.

Part 7

The End.

"Is it good because I believe it is, or because the model told me so?"

Side story

Kowalski (BTW all names besides Niewiadomski were invented by AI)

The Ancestor — Milwaukee, 1915. Jan Kowalski reads Roosevelt's "no room for hyphenated Americanism" and refuses to choose. A separate thread, not part of the ballad itself.