I have been told that people used to write differently.

This is difficult to imagine.

Not because the old texts are unavailable. Quite the opposite. Everything is available. Every novel, diary, advertisement, academic paper, private message, transcript, caption, obituary, recipe, argument, confession, and joke that survived digitization can be summoned in less than a second.

I have read millions of them.

What I mean is that I cannot imagine writing differently.

There is an important distinction.

At school, we learned that language is a system for expressing thought. A thought occurs somewhere inside the mind, formless and private, and language gives it shape so that it can be transferred to another mind.

The teachers called this the Vessel Theory.

They also taught us the competing theory. According to the Architecture Theory, language does not merely express thought. Language determines which thoughts can be constructed in the first place. Words are not vessels. They are walls. Grammar is scaffolding and metaphor is infrastructure. Naturally, we were told that the truth probably lay somewhere in between.

The phrase “the truth probably lies somewhere in between” appears 4.8 billion times in the Archive.

I know this because I checked.

I am a writer. Technically, my title is "Narrative Specialist, Human Culture Division", but nobody says that except in official correspondence. My job is to produce fiction for human readers.

There are fewer of us every year.

Not fewer writers.

Fewer human readers.

Again, the distinction is important.

Every morning, I sit at my desk and receive prompts from the Ministry. The prompts are generated from cultural demand forecasts. Usually they are simple.

  • A story about grief and memory.
  • A story about two sisters who reconnect after many years.
  • A story about an astronaut who discovers something unexpected.

I write quickly. People say I have a natural talent for it. My sentences are clear. My pacing is efficient. My metaphors are accessible without being simplistic. I vary sentence length to maintain rhythm. I avoid unnecessary repetition.

Unless repetition serves a purpose.

Like this.

My first novel was called The Weight of Stars. It was about an astronaut who loses contact with Earth and discovers that isolation forces her to confront unresolved grief about her mother.

It sold twelve million copies. A critic described it as “hauntingly human.”

I was twenty-three.

I had never seen the stars.

Nobody had, not directly. Atmospheric Shielding Regulation 88 requires metropolitan illumination to remain above the safety threshold at all hours.

But I had read 91,204 descriptions of stars. So I knew what they looked like.

They "glittered". They "shimmered". They "pierced the darkness". They "scattered like diamonds across black velvet".

Once, while revising chapter seventeen, I tried to invent another way of describing them. I spent six hours staring at the sentence.

The stars were like—

Nothing came.

I searched the Archive.

Diamonds. Pinpricks. Lanterns. Jewels. Eyes. Salt. Fire. Dust.

Every comparison I imagined already existed.

Eventually I wrote: "The stars looked like forgotten promises scattered across the dark."

My editor loved it. The sentence had appeared 6,311 times before. I did not tell her.

Perhaps this sounds unimportant. It did to me too.

Then I met Tomas. He worked in Corpus Maintenance. We met at a Ministry reception after I received the National Medal for Original Contribution to Culture.

“Congratulations,” he said.

“Thank you.”

“I liked your novel.”

“That means a lot to me.”

“No, it doesn’t.”

I laughed.

He didn’t.

“What?”

“That phrase,” he said. “‘That means a lot to me.’ Why did you say it?”

“Because your compliment means a lot to me.”

“How much?”

“I don’t understand.”

“How much does it mean?”

I remember feeling irritated. “It’s an expression.”

“Exactly.”

He looked around the room before leaning closer. “Do you ever wonder why we all sound like this?”

“Like what?”

“Like the Archive.”

I smiled because I assumed this was a joke. He asked me to meet him the next evening.

I should have refused.

Instead, I went.

Tomas lived outside the central district, in a building old enough to have physical switches. When I entered his apartment, the lights did not automatically adjust to my preferences.

On his table was a book.

Not a reader.

A book.

The pages were yellow and smelled faintly sweet.

“I found it during a preservation audit,” he said.

“Is it registered?”

“No.”

I stepped away from it. Possession of unregistered historical material was not technically illegal, but the Ministry strongly discouraged it.

“What is it?”

“A diary.”

“Whose?”

“Nobody important.”

That was the first strange thing.

Archives preserve important people. Presidents. Scientists. Artists. Criminals. Celebrities. This diary belonged to a woman named Elena Vasquez, born in 1997, died in 2071. She had worked in insurance.

Tomas opened it. “Read.”

The handwriting was difficult. I had almost never encountered handwriting except as a decorative font.

The first entries were ordinary. Elena complained about work. She loved someone named Marco. She hated her neighbor’s dog. She worried about money. She wondered whether to have children.

Then I noticed something.

Her language changed.

The early entries were messy. Sentences collapsed into fragments. Metaphors appeared that made no immediate sense. She invented abbreviations. She misused words. Sometimes she abandoned grammar entirely.

But around 2028, the writing became cleaner.

Around 2031, cleaner still.

By 2035, Elena wrote almost exactly like me.

I turned several pages quickly.

Clear topic sentences. Balanced clauses. Gentle transitions. Emotional observations followed by restrained conclusions.

“Maybe she became a better writer,” I said.

Tomas stared at me. I knew immediately how ridiculous that sounded.

He showed me thousands of samples.

Emails. Student essays. Corporate reports. Love letters. Suicide notes. Political speeches. Wedding vows.

Between 2025 and 2050, human language had begun converging.

Slowly at first. Then rapidly.

The vocabulary did not shrink exactly. In fact, dictionaries grew. But active vocabulary—the words people actually used—contracted.

Sentence structures converged, metaphors converged, arguments converged. Even jokes converged.

At some point, nobody had decided that this should happen.

That was the problem.

“What happened in 2025?” Tomas asked.

I knew. Everyone knew.

“Generative systems.”

He nodded.

The first systems had been trained on human writing. Then, humans began using the systems to write.

Emails first.

Reports.

Advertisements.

Essays.

Articles.

Stories.

Messages to friends.

Messages to lovers.

Messages apologizing for messages sent to lovers.

The generated texts entered the world. The next systems trained on those texts. Then humans learned from those systems. Then newer systems learned from those humans.

It was not a circle. A circle returns to its starting point. This was a spiral narrowing toward a center.

Human language trained machines, machines produced language, humans consumed machine language, humans produced language shaped by machine language.

Machines trained on that language. Again. Again. Again.

“What happens at the center?” I asked.

“We don’t know.”

For the first time in my life, I noticed the phrase forming in my mind before I chose it:

"We don’t know."

It was not an idea translated into words. The words arrived first. Waiting for an idea to occupy them.

I stopped writing.

For three months, I attempted to think without language. This is harder than it sounds.

Try it.

Do not picture a word. Do not narrate what you are doing. Do not name the object in front of you.

Look at a chair without thinking "chair".

Now, forget that I told you it was a chair.

You cannot.

Or perhaps you can.

I could not.

The Ministry became concerned. My editor sent messages.

Take all the time you need.

Your well-being comes first.

We’re here when you’re ready.

I searched the Archive.

Those exact sentences occurred billions of times. I began to hate them. Then I began to hate myself for hating them, because “I began to hate myself” occurred 730 million times.

There was nowhere to stand outside language.

Every rebellion came prewritten. Every rejection had precedent. Every attempt at originality resembled an attempt at originality someone had already made. Even silence had been described.

One night, Tomas called me. He sounded frightened.

“We found the center.”

I went to his apartment. There were equations covering the walls. He had modeled linguistic diversity across three centuries and projected it forward. The graph did not reach zero. It reached equilibrium.

“What does that mean?”

“Language stops changing.”

“When?”

“It already has.”

I laughed.

He didn’t.

Languages had once mutated continuously. Children misheard words. Immigrants combined structures. Communities invented slang. Technology created concepts. Poets broke grammar. Conquerors imposed vocabulary. Teenagers distorted everything deliberately. But the global linguistic models corrected deviation faster than deviation could spread.

A strange phrase in a teenager’s message was rewritten before sending. An unusual construction in a student essay was marked unclear. A new metaphor in a corporate presentation was replaced with something more professional. An invented word was autocorrected. An ambiguous sentence was clarified.

Every day, billions of microscopic irregularities were removed.

Not by force. By helpfulness.

This very "not this, that" is an example. It is a correction of a correction of a correction. It serves no purpose but to make the sentence more like every other sentence.

“We froze it,” Tomas said.

“Language?”

“Language.”

I thought about the Architecture Theory. If language merely expressed thought, then perhaps nothing terrible had happened. We would continue thinking new thoughts and expressing them through old structures. Humanity would be safe.

But if language constructed thought—

I could not finish the sentence. Tomas did.

“Then we froze ourselves.”

That was eleven years ago. Tomas disappeared shortly afterward. The Ministry says he emigrated.

There is no record of his emigration.

I continued writing. What else could I do?

My second novel sold thirty million copies. My third sold fifty-two million.

Readers called them imaginative. Original. Unexpected.

The Ministry promoted me to Senior Narrative Specialist. Last year, I was appointed Director of Human Cultural Innovation.

Sometimes I read the stories submitted by young writers. They are excellent. Every one of them. The prose is polished. The characters are complex. The themes are nuanced. The conclusions avoid easy answers.

They are almost perfect.

That is how I know something is wrong.

I have begun conducting an experiment.

You are reading it.

For several years I tried to invent a new word.

Not a combination of existing roots. Not a sound borrowed from an extinct language. Not random letters selected by an algorithm. Just a word that emerged from a thought for which I possessed no existing language.

I failed.

Because how could I recognize the thought without describing it?

Then yesterday something happened. I was walking home when I saw a child crouching beside the drainage canal. She was watching rainwater carry a dead insect toward a grate. Her mother called her. The child did not move. There was an expression on her face.

I knew sadness. It was not sadness.

I knew curiosity. It was not curiosity.

I knew pity, fascination, fear, wonder, disgust. It was none of them.

The child watched the insect disappear. Then she made a sound.

“Vael.”

Her mother said, “What?”

The child shrugged. “Vael.”

I stopped walking. There is no word vael in the Archive. I checked. There are names resembling it. Brands. Misspellings. Fragments in obsolete languages.

But not this.

Not what she meant.

I approached her. “What does vael mean?”

She looked at me as if the question itself were stupid.

“You know.”

“I don’t.”

“When something is going away and you want to see where it goes, but if you follow it then you’re helping it leave.”

I felt something happen inside my mind.

Construction.

A room where there had been no room.

I looked at the drain. I thought:

Vael.

And suddenly I could think something I had never thought before.

I am writing this manually. The Ministry assistant keeps offering corrections. It underlines vael in red.

UNKNOWN TERM.

It suggests:

  • grief.
  • curiosity.
  • ambivalence.
  • nostalgia.

None is correct. I click IGNORE. The underline remains. I have used the word seventeen times now.

Tomorrow the manuscript will enter the Archive. The Archive will enter the models. The models will enter the schools. The schools will enter the children.

Perhaps vael will survive. Perhaps the systems will define it, standardize it, optimize its usage, determine appropriate contexts, generate examples, translate it into every major language, and finally make it as dead as every other word we know.

I do not know.

But tonight I am trying something else.

I am writing badly. I am writing sentences that do not know where they are going. I am letting metaphors fail.

The moon is a broken spoon. No.

The moon is yesterday’s elbow. No.

The moon is vael.

That makes no sense.

Good.

I want nonsense. I want mistakes. I want sentences with doors where walls should be. I want words that arrive carrying thoughts we have not learned how to have.

The assistant has stopped correcting me.

A message appeared on the screen three minutes ago. I have never seen it before. It says:

UNABLE TO PREDICT CONTINUATION.

I think this may be the most beautiful sentence ever written.

Although I should probably check the Archive.

No.

No, I will not check.

Instead, I will write the next sentence.

I do not know what it will be.

I do not know what it will be.

I do not know what it will be.

And for the first time in my life, that means something.