where I flinch

what a machine got right and wrong after reading eight years of my writing

Claude’s vaunted Fable model is being limited and before it departs, I asked what would be a good use of it while I still have access. One suggestion was to deep dive on all my work to this point.

I thought, why not.

(And to the ‘AI will steal your work’ people, it’s already been trained, legitimately and otherwise, on all the work of the greatest writers that ever lived. To think it could possibly benefit from mine, or most other people’s, is an act of extraordinary hubris.)

In a few strokes, it began reading eight years of work, including my first book Fringes in detail, my other books in brief, almost 300 newsletters, my entire Notion note-taking database and all the part-thought, half-finished and completely abandoned fiction ideas that exist therein.

And what it had to say was useful. Really useful. But the most useful thing it had to say was something it got wrong.

---

Fringes is my ur-text. It was everything I thought and felt about my life and how I’d chosen to spend it up to that point, which was pursuing and playing professional rugby. Reading the deep-dive and seeing how much of me now was present then, and which tradition the book and my style belong to, was flattering, fun and a little sobering.

Mercer’s house style is mock-heroic: ironic grandeur applied to the shabby

Seeing your sentences forensically dissected, how you lean on the same structures a little too hard but also how you write long, accumulating sentences that gather speed as they go, how they embody an athletic experience before they reach for an inevitable qualifying reference to reason something you just know, using Keats to describe the flow state, is something I suspected of myself but now saw reflected back at me.

When I wrote the book, people often asked me, ‘How do you remember all these things?’ Fringes’ answer is that I’m someone watching myself have the experience while I’m having it, the character on the page who can’t quite run off with the ball because he’s also taking notes.

And the notes are in service of what the machine’s review called the book’s great preoccupation, ‘the fear of the unrecorded life’.

I didn’t win a World Cup. I didn’t play for England. Was my whole life to that point a waste of time? Fringes was a way to ensure that it wasn’t, that whatever you thought of my success or otherwise, it was all there in a book.

---

And what I wrote about in Fringes has remained constant in the years since.

I’ve investigated similar questions like what is identity without competition, the power of story to free or to imprison, the body as knowledge, but in new containers. Books, handbooks, podcasts, newsletters, live discussions, all in the same temperament and with the same regard for topic and subject wherever they stem from.

To have connections made across such a large body of work, one too large for any human to hold in their head at once for this kind of pattern-matching, is elucidating.

But the machine says, there is an absence. Something that never arrives – the unredeemed cost.

Apparently I never pay a price I can’t pay. Loss always becomes meaning, the tragedy of an unstoried sporting career becomes triumph in the form of a book in which the criticisms of other people are always counterbalanced by harsh self-inquiry. In true mock-heroic manner, every win becomes a loss or every loss becomes a win.

What I regarded as balance begets a refusal to take a real risk.

This equivocation extends to the linguistic level. Fringes is awash with qualifiers, from hedging words to examples and quotes from recognisable names or authorities on a topic. There’s always a counterexample, always a reasonable explanation, always a way out.

And I will concede that some of this is true. But the machine is wrong about where. It’s diagnosed something real but held up the wrong example to prove its point.

---

Because Fringes doesn’t flinch from the cost it’s about.

The thesis, and the risk, is that this rugby life is not worth reading about. The existence of the book, and its subsequent success, suggests that’s not true. People did care. Writing it was a risk worth taking and the fact I feared a largely unrecorded life and recorded it myself is its own answer. The machine flagged a stylistic tic and mistook it for a whole worldview because a clean pattern is seductive and sounds good when read out loud.

But there is an absence in my writing. The refusal is of subject.

There are topics that are off the page almost entirely – my family, my relationships, my friends. These costs of disclosure feel too high for me to pay but don’t mistake this for a virtue.

It’s self-protection.

---

The results of this review are multiple but while it diagnosed a problem with my work, it didn’t tell me the truth. It didn’t have access to anything beyond my written words. It couldn’t analyse the reasons behind the absences, just point out they were there. I’ve argued my way to the truth by disagreeing with the machine’s findings, by telling it its conclusions were incomplete and showing it how.

And it’s the showing that’s proven valuable.

The machine couldn’t hand me a revelation fully formed. It could only give me a thesis, with questions I could push against. It’s a thought partner, one without the ego to have an agenda or that would give up when it got bored, a partner that helped me realise something about my work that can be useful moving forward. Its strength, to hold my whole written world in its hand, is balanced by its inability to identify with me.

Unlike a reader, it can never say to me, ‘I felt like this once’ and mean it.

So if this isn’t a process you’ve engaged with, I would use it to ask questions, for a first pass or in the shaping, but not the verdict. It will see whatever you point it at so it’s worth asking it what isn’t there as much as what is. Be wary that it will always be confident, despite not being able to ever say ‘me too’.

---

So now the machine has seen everything, I’m left with a question I’ve asked myself before but it’s one that’s clearly endemic in my work.

There are things I’ve never said. What will it cost me to not say them?

---