I. I thought AI might finally give my thoughts somewhere to go
I have wanted to write for a long time.
Not necessarily to become a writer, at least not in the way I understood the word. I wanted somewhere to put the thoughts I was already having: things I noticed, questions I kept returning to, ideas I wanted to follow far enough to see where they went. Some I wanted to share. Some I simply wanted to keep.
The problem was that wanting to put those thoughts into words and actually doing it were two different things.
Give me a thought and I can talk about it for hours. I can question it, contradict myself, make connections, remember something seemingly unrelated, change my mind halfway through, and come out somewhere I could not have predicted when I started.
Give me a blank page and ask me to write the same thought down, and sometimes there is suddenly nothing there.
The thoughts existed. The bottleneck was getting them into a form that could survive outside my head.
Then I started talking to AI.
At first, I did not think of what I was doing as writing. I was talking. I would bring a thought, an idea, a reflection, sometimes something half-formed that I was still trying to understand. The AI would respond. I would tell it what it had misunderstood, which part interested me, or why something it said made me think of something else. It would build on that. I would build on its response.
Somewhere inside those exchanges, things I had struggled to put on a blank page were becoming articulate.
More than that, often the thought itself became better.
A distinction I had not made became visible. A question became more precise. An assumption got challenged. Something I had felt but could not name acquired language, and once it had language, I could examine it, disagree with it, reshape it, take it somewhere else.
That realization made me unexpectedly hopeful.
Maybe this was a way for me to start putting my thoughts into words and, eventually, putting them out there.
Maybe writing did not have to begin with me sitting alone and producing finished sentences from nothing. Maybe conversation could be one of the ways I arrived at them.
For a moment, that felt like a possibility.
Then, the next day, I came across an AI writing policy.
I agreed with much of what it was trying to protect. Writing is part of thinking. If you put your name on something, you should understand it and be able to stand behind it. AI can assist the process, but it should not become a substitute for authorship.
I remember feeling that possibility from the night before start to close.
I did not want to ask a machine for an argument I had never thought through, make a few cosmetic changes, and call the result mine. If I published something under my name, I wanted to understand what it meant. I wanted to know what I believed, where I was uncertain, and why I had chosen to say what I was saying.
But I kept thinking: it can't be just that.
The process I had discovered did not feel like the disappearance of my thinking. If anything, I was thinking more. I could stay with an idea longer because there was something responding to it, challenging it, or giving me another direction to follow.
Going back to the blank page did not seem like a more ethical solution. It just meant going back to an obstacle I had not yet learned to overcome.
So I started wondering whether I needed another answer.
What if, instead of choosing between writing alone and giving the writing over to AI, I created some rules for how the collaboration should work?
A kind of writing governance.
Something that would let me use the process that was helping me while protecting the things I did not want to give away: my judgment, my responsibility for what I publish, and whatever makes the work meaningfully mine.
I did not know what those rules should be yet.
I still don't entirely know.
But that question felt much more useful than asking whether AI had been involved at all.
II. What exactly am I trying to protect?
There is a version of AI-assisted writing that makes me uncomfortable too.
A model can produce a coherent argument in seconds. The reasoning may even be good. But if the person putting their name on it has never really wrestled with what the piece is saying, something important seems to be missing.
That is not the kind of writing I want for myself.
If I publish something, I want to understand the argument well enough to challenge it. I want to notice when a sentence sounds convincing but does not actually say what I believe. I want to know where I am uncertain.
So I understand the instinct to draw a line.
The difficulty is deciding where that line belongs.
Writing every sentence yourself is evidence of participation. But it is not, by itself, proof that you have wrestled deeply with the ideas behind them.
And the presence of AI does not automatically prove the opposite.
A person can write something shallow without AI. A person can also think deeply in conversation with it.
That made me wonder whether I was focusing too much on who produced the sentences and not enough on what happened in the thinking.
Maybe what needs protecting is not solitary sentence production.
Maybe what needs protecting is the human being's continued presence in the work.
But even that becomes difficult to define.
Is it enough to agree with what an AI gives you?
I don't think so.
Is editing enough?
Not necessarily.
What if you reject part of an argument, change the premise, ask another question, follow a connection you had not seen before, and eventually arrive somewhere neither the original question nor the first response contained?
That feels different to me.
Not because the AI contribution has disappeared.
Because mine hasn't.
III. The collaboration is real. So where does authorship live?
One reason this process is valuable to me is that it compresses the loop around an idea.
Before AI, following a thought could mean repeatedly leaving it: looking for sources, reading competing explanations, finding evidence that supported or challenged what I was thinking, then returning to the original question with new information.
I still think those sources matter. Anything I eventually publish as fact still has to be verified against sources appropriate to the claim.
But conversation with AI makes the exploratory part much faster.
I can encounter another interpretation, a counterargument, a concept, or a possible connection without losing the thread of the thing I was trying to understand.
That does not mean I arrive at whatever answer the AI gives me.
It means I can move through more possibilities.
We can wrestle with competing ideas until I can see which pieces connect, which ones do not, and which ones make the original thought more coherent to me.
Sometimes I start with one question and end somewhere else entirely.
I may set the direction I am trying to follow, but I do not always know the destination in advance.
That is part of the point.
The AI can materially affect the path. It can introduce something I had not considered or give language to something I had sensed but had not articulated.
I do not want to pretend those contributions do not exist simply because acknowledging them makes authorship harder to describe.
But I am still deciding what matters, what I believe, what I reject, what needs verification, and what I am willing to put my name on.
And even that may not be a complete definition of authorship.
While working on this essay, I encountered a useful boundary myself. There were passages I agreed with completely but still felt were not quite mine. The ideas were right, but the framing felt too much like the AI's way of making the argument.
Agreement, I realized, is not enough either.
So I am left with a question I have not completely answered:
If both sides can materially shape the work, where exactly does authorship live?
IV. Governance instead of purity
I don't want figuring all of this out to become another reason not to put my thoughts out there.
I don't expect this to be the only way I ever do it. Maybe I will become more comfortable facing a blank page. Maybe publishing regularly will itself change how I write and how I think.
But right now, I have thoughts I want to put out there, and I have found a process that helps me do that.
So rather than waiting until I have solved the question of AI and authorship, I am trying to develop boundaries I can work with and let the practice show me where those boundaries need to change.
For now, that means understanding what I publish. Knowing what I actually believe and where I am uncertain. Making meaningful decisions about what stays, what goes, and how the argument takes shape. Verifying claims that need verification. Taking responsibility for what I put my name on.
And if AI has materially shaped the thinking or the writing, I would rather acknowledge that than pretend the work emerged from me alone.
That still leaves questions.
How much AI-generated language can remain before the work begins to feel less like mine? When does collaboration become delegation? Is authorship located in the origin of an idea, the production of language, the judgment exercised over a piece, or some combination of all three?
I do not have fixed answers yet.
That is partly why I want to publish this.
Until now, much of the feedback loop around these thoughts has happened between me and AI. Once they are public, other people can participate in it.
Someone might recognize themselves in this. Someone else might completely reject the boundary I am drawing. Someone might notice an assumption I have missed, draw a different line for themselves, or take the question somewhere I would never have taken it.
Maybe publishing becomes another way of thinking.
Maybe the feedback loop becomes larger than the one I have with AI.
And maybe, over time, that changes the way I write too.
For now, I want to start.
Not because I have resolved what authorship means in this new arrangement.
Because I have something I want to think about with other people.