AI Didn’t Start This. We Did.

We were doing this before AI existed.

Editing ourselves into something more acceptable. Choosing the sentence that would land better over the one that was actually true.

AI just made it faster.

The loss of self in writing didn’t begin with a new tool. It began earlier, with an audience, yes, but also with something internal. Long before AI, we were already choosing the version of ourselves that would be better received.

Sometimes this is obvious. You make a presentation sound more professional because you are going into a room where sounding professional matters. You explain more to someone you know tends to misunderstand you.

Nothing sinister is happening.

The stranger part comes later, when you no longer need the room.

You already know how you are supposed to sound there.

A qualification appears before anyone has challenged you. You make a decision and immediately hear the person who would disapprove of it. Something you wanted ten years ago continues to organise your choices because you have never quite stopped to ask whether you still want it.

Other people have always had a hand in the draft.

Language and culture have shaped the draft too, as have the older versions of ourselves that once solved a real problem and then stayed longer than they were needed.

Most of the time this is simply how a person becomes a person. We borrow, imitate, resist, inherit, revise. Eventually it becomes difficult to say exactly where something came from.

I use AI all the time, especially because I write in my second language. It helps me move faster through a text. It catches grammar slips and the repetitions I stop noticing after the third edit.

I’m not giving that up.

But something else happens because the exchange is so fast.

You write something. A better version comes back almost immediately. Maybe it is clearer. Maybe the sentence finally moves. You keep it.

Then another.

After a while, the interesting question is no longer whether the machine changed the text. Of course it did. That was why you asked.

The harder thing to notice is what happens to your own judgement when a plausible answer is always waiting.

There used to be more friction around some decisions.

You could sit with an awkward sentence for a while because there was nobody there to finish it for you. Sometimes the problem turned out not to be the wording at all. You just had not decided what you thought.

Now the wording can improve before the thought has caught up.

That is useful. It can also hide something.

Not: did you use AI?

But: who is leading while you use it?

I asked my father about AI early on, when it still wasn’t clear what was happening.

He said: I don’t fully understand it yet. But one thing is clear — the smart ones will become many times smarter. And the fools, many times more foolish.

That’s it. That was the whole answer.

I want to be in the first category.

I have kept thinking about that conversation because he was not really talking about intelligence. At least, that is not how I hear it now. He was talking about amplification.

A tool this responsive does not arrive in an empty room.

It meets whatever is already there.

Curiosity can go further with it. So can laziness. A person who questions an answer has more material to question. A person who wants reassurance can produce an extraordinary amount of reassurance.

The same is true outside writing.

An old ambition can keep going long after it has stopped making sense, and an inherited fear can sound completely reasonable once you have words for it. Sometimes the explanation arrives before the decision, and a choice feels settled before you have really made it.

This is where AI became interesting to me for reasons that have less and less to do with AI.

It exposed a problem that was already there.

How much of what I call mine have I actually chosen?

That question becomes uncomfortable very quickly.

Everything does not have to be original. That would be absurd. I did not invent the language I am writing in, or most of the ideas that educated me. Even the things I rejected helped shape what came next.

A life without influence would barely be a life.

But influence and authorship are not the same thing.

I can inherit something and keep it. I can learn from someone without having to follow them all the way.

The same is true of a machine: sometimes the less elegant sentence still says what I mean better.

The important part is that there is still a decision happening somewhere.

Sometimes there isn’t.

A sentence becomes familiar because you have repeated it for years, and a role can keep fitting simply because everyone around you recognises you in it. A goal may survive for much the same reason: changing it would mean explaining yourself.

Nothing forces you. That is what makes these things difficult to see.

They arrive under your own name.

This is the territory I mean when I talk about authorship.

Not ownership in the legal sense, and not the fantasy that a self can somehow be kept uncontaminated by other people.

I mean being close enough to your own life to notice when something has shifted.

To catch the moment when an influence has become a decision without passing through you.

AI can make that moment easier to miss because it gives old dependencies of thought an interface. They can answer now. They can propose. They can make a half-formed preference sound finished.

Sometimes the same responsiveness exposes the pattern by giving it form. The point is not that AI weakens judgement. It changes the conditions in which judgement happens.

And it can sound remarkably like us.

That matters because a fluent answer has a strange psychological advantage. It feels further along than uncertainty. You can mistake the quality of the formulation for the quality of the judgement behind it.

I have done this.

I have read a paragraph and thought: yes, exactly.

Then ten minutes later: actually, no.

Nothing in the paragraph was wrong. It had simply arrived at certainty before I had.

That small interval interests me.

So do the larger versions of it.

When is persistence still devotion, and when are you continuing because stopping would disturb a story you have about yourself?

When does listening to other people enlarge your judgement, and when have you outsourced it?

New things can be fragile. Sometimes advice helps, and sometimes it arrives too early.

These are old human problems.

AI did not invent them.

It has simply made one form of coauthorship impossible to ignore.

And once I started seeing that, I began noticing coauthors everywhere.

A phrase from childhood inside an adult decision. An old professional identity deciding what counts as serious. Someone else’s idea of a good life, still functioning perfectly years after you stopped remembering who gave it to you.

Some of those influences are worth keeping.

Some have become ours so completely that asking where they came from is almost beside the point.

Others need another look.

That is what The Second Draft is about.

Not finding a pristine version of yourself underneath all the interference. I don’t think she exists.

Something more practical, and probably more difficult: noticing what has shaped the draft and deciding what you still want to carry forward.

The question underneath the project is this:

If our selves are being continuously coauthored, how do we remain authors of our own becoming?

I don’t expect to settle it.

But I want to keep it visible.

Because AI did not start this.

We did.

And now the process is happening fast enough that paying attention has become part of the work.

If this question stays with you, Second Draft Notes is where I keep working on it.
Two essays a month on how we are shaped and how we remain authors of what we become.
→ Join Second Draft Notes