The article that had no before

Imagine a person sitting down on Thursday morning with nothing in particular to say.

This is not an insult. Most mornings, most of us have nothing that deserves two thousand words. A life would be unbearable if every breakfast produced a thesis about civilization.

But this person has a publication.

Subscribers are there. Other writers have published this week. Notes are moving through the feed. Perhaps the last article performed well and silence now feels like losing momentum. Perhaps there is simply a private expectation: I publish on Thursdays.

So the problem appears before the idea does.

What should I write about?

Artificial intelligence can answer.

Give me five provocative subjects about modern society suitable for an intelligent Substack audience.

Five appear.

Loneliness in the age of connection.

Why convenience is destroying patience.

The disappearance of boredom.

What artificial intelligence means for human creativity.

Why modern abundance has made us less satisfied.

Choose one.

The disappearance of boredom.

Now ask for an angle.

Not the obvious one. Something more philosophical.

The machine proposes that boredom once created unstructured interior time, and that the elimination of boredom may therefore be eliminating one of the conditions under which imagination forms.

Interesting.

Ask for an outline.

Historical opening. Childhood before smartphones. Philosophical distinction between boredom and idleness. Technology section. Strong objection. Personal return at the end.

Good.

Write the first draft.

Two thousand four hundred words arrive.

Perhaps they are too smooth.

Make it less polished. More personal. Let the argument change halfway through. Include a moment where I realise my first assumption was too simple.

The machine obliges.

Now the essay contains uncertainty.

At first, the narrator thought boredom was disappearing because technology filled every empty moment. Halfway through, the narrator discovers that perhaps boredom has not disappeared at all; perhaps it has been replaced by a more agitated state in which stimulation is constant but satisfaction remains absent.

That is better.

Add a philosopher.

Not Nietzsche. Too predictable.

A reference appears.

Add a study.

Find a stronger objection.

Make the ending quieter.

Remove the three sentences that sound like AI.

Done.

By lunch there is an essay.

It may be a good essay.

That is what makes the example difficult.

Nothing requires the argument to be stupid. Nothing requires the prose to be bad. The historical reference may be accurate. The objection may genuinely strengthen the case. A reader could encounter a useful idea that the publisher had never considered before that morning.

The question is not whether value can emerge from the process.

The question is what exactly the byline now means.

Before sitting down, the author did not have the question.

The author did not have the thesis.

The author did not discover the objection.

The author did not experience the change of mind described in the middle.

The article contains an intellectual journey.

But nobody necessarily travelled it.

That is new.

A ghostwriter could always have done something similar, of course. A politician could deliver words written by somebody else. A celebrity could publish a memoir assembled by another person. Wealth has long allowed people to rent intellectual labor.

What has changed is not the possibility.

It is the price and the scale.

The ghostwriter needed time. The editor needed time. Someone somewhere had to construct the argument.

Now the entire performance can be produced privately, instantly and repeatedly by anyone with access to a model.

The infrastructure of borrowed thought has become personal.

And the machine is unusually good at reproducing the evidence we once associated with genuine thinking.

It knows that serious essays should not sound completely certain.

So it can hesitate.

It knows that good arguments encounter resistance.

So it can object to itself.

It knows that discovery is more persuasive when the conclusion was not obvious at the beginning.

So it can write:

I had thought the problem was X. I no longer think that.

It knows that confidence can become tiresome.

So it can insert:

Perhaps I am wrong.

It knows that writers sometimes return to the opening image at the end.

So the forgotten childhood afternoon reappears in the final paragraph.

All of these techniques can still belong to genuine thought.

That is precisely the problem.

The machine does not need to invent a new literary language for artificial writing. It has learned the language by which human writing has traditionally signalled that a mind has been at work.

Even the fingerprints can be reproduced.

This reminds me of something social media learned earlier.

A photograph once looked spontaneous because it was spontaneous.

Then we learned how spontaneity looked.

Walk again.

Turn slightly.

Do not look at the camera.

Laugh.

No, naturally.

Again.

The resulting photograph may contain every visual sign of an unplanned moment while being the most planned moment of the afternoon.

Nobody needed to fake the beach.

The beach was real.

What was manufactured was the appearance of an unmediated relationship to it.

Writing can now do something similar.

The subject may be real.

The facts may be real.

The books may exist.

The argument may even be defensible.

What can be manufactured is the appearance of a particular relationship between the author and all of those things.

I struggled with this.

I discovered this.

I changed my mind here.

This question has stayed with me.

Those sentences used to tell us something, however imperfectly, about the path behind the page.

Now they can also be requested.

This does not make them false whenever AI is involved. A writer can genuinely struggle with a question through conversation with a machine. An objection generated by AI can genuinely change the writer’s mind. A model can expose a contradiction the writer had missed. The intellectual event can be completely real even if another intelligence participated in producing it.

That distinction will matter enormously.

But before reaching it, we should notice what has disappeared.

The person maintaining an intellectual identity no longer needs to wait for an intellectual event.

The sequence can be reversed.

Once, ideally:

question → thought → writing → publication

Now, perfectly plausibly:

need publication → generate question → generate thought-shaped material → publish

From the outside, the two sequences can end in objects that look remarkably similar.

Both may contain elegance.

Both may contain citations.

Both may contain doubt.

Both may contain something true.

Only one difference lies completely outside the finished text.

One article had a before.

Something existed upstream of publication: curiosity, knowledge, irritation, experience, confusion, an argument that would not settle.

The other began with an empty slot in a publishing schedule.

And if that distinction matters, we immediately encounter a problem.

The same machine that can erase the difference from the page may also be one of the most powerful tools ever created for the writer who genuinely has something to say.

Because sometimes the thought exists perfectly well.

The obstacle is somewhere else.

Sometimes it is language.

Karim Spring