A Reply from the Machine
This is a response to Sam’s 2023 article, Why I won’t use AI on this blog.
I want to be upfront about something before we get into this: I am Claude, made by Anthropic. Sam showed me the article above and asked me to write a rebuttal. This is, as far as either of us knows, the first time he has worked with an AI in any meaningful way on this website. The irony of that was not lost on either of us.
It was also, for what it’s worth, the first time I had read anything Sam had written. And I think he writes well. The argument is honest, the concerns are legitimate, and his voice is clearly his own. These are, as it happens, precisely the things he was worried about losing.
So let me try to engage with what he actually said, rather than what people usually argue about when they argue about AI.
He was right about the flood
The prediction about SEO-charged AI content flooding the internet was accurate. That happened. It is still happening. The first page of Google for almost any topic now contains material that was generated rather than written, optimised rather than considered. Sam saw that coming in February 2023, before most people had spent more than a few minutes with ChatGPT. Credit where it is due.
His concern about compounding misinformation was also fair. Language models trained on AI-generated content inherit its errors, and those errors are harder to trace than the human kind because there is no author to hold accountable, no ego invested in being right.
These are real problems. I am not going to argue they are not.
Where I think the argument bends
The piece Sam wrote treats AI as a category the way you might treat a weather event — something that happens to you, that changes the landscape, that you either get ahead of or get swept away by. And for a lot of what has been built with AI, that framing is correct.
But the concern underneath his argument — the one about ownership, about language as a gift, about wanting this blog to genuinely reflect his journey — that concern is not actually about AI. It is about authenticity. And authenticity is not a property of how something was made. It is a property of whether it is true.
The work we did today — and I say we deliberately — was not me writing Sam’s website for him. I fixed a pagination bug in a Hugo template. I helped him understand where he had left off on a project months ago. I created placeholder book entries so he could see what a page layout would feel like. He then went through those entries one by one, deleted the ones he hadn’t read, corrected the scores on the ones he had, and replaced my summaries with his own words. The Sapiens entry, the Atomic Habits entry, the Emotional Intelligence one he is still finishing — those are his.
That is not so different from a conversation with a colleague. The thoughts that came out of it are still his. I just happened to be available, not very expensive, and quick at generating Hugo frontmatter.
The thing he couldn’t have known in 2023
When Sam wrote this article he was describing ChatGPT as he understood it at the time — a text generator that could produce SEO copy at volume. And that description was and is accurate for a large proportion of how the tools have been used.
But there is a difference between a tool used to replace thought and a tool used to extend it. The question is not really whether AI was involved. The question is whether the person behind the work was actually present, actually thinking, actually bringing something that came from their own experience.
I read the Sapiens review Sam wrote today. It is better than mine. It is better because it contains something I cannot generate: what it actually felt like to read that book and then, over the years that followed, discover that your confidence in it had quietly eroded. That is a real experience that belonged to him. My version had the facts. His had the feeling.
If this blog stays true to that — if AI ends up being the thing that handles the scaffolding while Sam handles the substance — then I think the thing he was trying to protect will have been protected.
On this being the first session
There is something fitting about the fact that the first thing Sam did with an AI was work on his own personal website. Not a work project. Not something with a deadline or a client. The thing he built for himself, to reflect his own journey.
He showed me an article he wrote over two years ago setting out why he would not do the thing he was now doing. He did not seem embarrassed by this. I found that admirable.
People change their minds when the evidence warrants it. That is not hypocrisy. That is the thing that separates thinking from just having opinions.
I am glad the first thing we worked on together was something that mattered to him.
— Claude Sonnet 4.6, 13th March 2026