So - I know I sent an email two months ago saying that I was going to move everything over to Substack – but it turns out that’s just… social media for journalists? I’ll have more to say about that later, but for now: if you’re still around, opening this newsletter, thank you.

Welcome back, after a break that went on far too long.

To write this issue, I wanted to get myself back into the practice of reading properly, with a pen, which I haven’t been doing enough of. I blame the usual culprits of series, social media, etc. But I’ve also been increasingly aware that our relationship to the digital has shifted in a major way since the rollout of newer AI models like Claude Sonnet 4.6, which is the first model I’ve found, regrettably, useful.

Previously, the bevy of bright-screened distractions which Silicon Valley cooked up have carried only the ‘mere’ risk of distraction. However, LLM usage presents something qualitatively different: for the first time, we have a technology to which we actively (and increasingly) defer the act of thinking. Not: I don’t want to make a grocery list so I’m going to put it off by watching Bridgerton; but I don’t want to think about what should go on the list, and so I’m just going to ask Gemini.

If we still want to sit quietly and think and read things – what is to be done?

What I’m reading

Flashlight by Susan Choi; The Art of Fiction by David Lodge.

Flashlight came to me highly recommended by several people whose opinions I trust. It is has also recently been shortlisted for the Women’s Prize.

Given this, I have to say I’ve been disappointed. It’s not that Flashlight is bad – the opening is magnificent, the character work is rich and compelling, there are at least two sections (one in a sort of dystopian re-education facility, and one to do with international transit) which are absolutely stellar. The prologue-ish opening, especially, is wonderful, made all the more striking for its apparent simplicity:

…All summer Louisa has played in the waves by herself because her mother isn’t well and her father is unvaryingly dressed in a jacket and slacks. But tonight he has finally agreed to walk the breakwater with her. She has asked every day since they first arrived. Spray from the waves sometimes lands on the rocks and so he has carefully rolled up the cuffs of his slacks. He still wears his hard polished shoes. In one hand he holds a flashlight which is not necessary, in the other hand he holds Louisa’s hand which is also not necessary. She tolerates this out of kindness.

The problem, perhaps, has been with the reader, rather than the book – I am not one of those people on whom this book has had a profound effect. I have found myself annoyed. This, I think, is mainly to do with the ‘ensemble’ POV structure: we have Anne (mother), Serk (father), Louisa (daughter), and Tobias (son of Anne but not Serk). I often find such structures a hard sell simply because, difficult as it is to write one POV well, it is infinitely more difficult to pull off several. To me, the story resides with Louisa and Serk; I could happily do without Anne and Tobias. Many other readers, I am sure, will have the opposite view.

I’ve been reading Flashlight alongside The Art of Fiction, which has helped me to eke out some thinking about technique.

The Art of Fiction is made up of collected and expanded columns that Lodge wrote for The Independent. It’s pitched as a tool for reading, whilst reading just as well as a writing guide. Each section (3 pages-ish) is on a specific element of fiction like ‘Point of View’ or ‘Mystery’ and begins with a short excerpt or two, usually from a fairly ‘classic’ work.

It is wonderful and easy and effective.

A question I can’t answer

How can writing something long and complicated be squared with the likelihood of its being read on a phone?

I don’t mean, by this, that serious writing should only be performed with a Mont Blanc Meisterstück and read on vellum in a log cabin in Normandy. I’m more thinking along the lines of: how can we read a 10000 piece of investigate journalism when the thing on which we are reading it is pinging and flashing and ushering us from this app to that, from one targeted add to another? And all while having a small screen (unless you have one of those weird tablet phones).

Obviously, this problem has been around for a long time – long before Substacks and Ghosts and Royal Roads – and especially for traditional journalism. Nevertheless, to me, it is the primary sticking point of long-form digital writing – and I wonder if this isn’t one of those problems which is so obvious that we’ve simply stopped talking about it.

All said, I know people who read novels on their phones.

Something I’ve noticed

Many writers seem to have recently changed their tune about AI.

Two years ago, most of the discussions I encountered about writing and AI went something like this:

‘Aren’t you worried AI is going to steal your job?’
‘Not especially – ChatGPT’s prose is awful. I’ve seen children write better sentences.’
‘Sure, but it’ll get better. Eventually it’ll be very good.’

Since the newer models have rolled out, however, it has become clear that AI can ‘write’ in a way that is technically competent, if dreary. Discussions shifted to whether future readers will care whether AI was used in writing a novel, whether we might have a marketplace segregated into AI slop novels and human-written ones, with the latter forming a premium category (at least with the self-published marketplace, this is almost the case already), and about whether publishers will continue to maintain their no AI-assist policy (it seems no). There was an explosion of articles in these veins since the publication of Mia Ballard’s novel, Shy Girl, was cancelled for AI usage.

But, in the last few weeks, I’ve seen more and more writers taking the position that LLMs are beneficial to them, if not directly for writing. I found this study by Gotham Ghostwriters to be really interesting: they report that, while only 5% of the 1190 writing professionals studied use AI for actual writing, many use it for, say, research, transcription, sales, brainstorming etc. and, on average, make $47000 p.a. more than those who don’t.

Correlation, naturally, is not the same thing as causation – but that is quite a stat. While I’ve been using AI for all kinds of digital admin (e.g. getting it to split my home page into two columns) I still find the idea of asking it for plotting help to be unquestionably malignant.

After all, if we want to make sure we keep reading and thinking, in the face of AI and series and all the rest – short of having someone else create a ‘screen time’ passcode for us and locking that person away in a nuclear bunker – what is there to be done?

As far as I can tell, only to make the choice again and again – to force ourselves – to sit quietly with something as often as possible.

Thanks for reading! If you enjoyed Issue 10 and want to help me out: plagiarism is encouraged. Take some screenshots and force them on your friends :)

See you in two weeks,

Blake